Title: Data Unchained Episode 13
Molly Presley: Hello, and welcome to this episode of Data Unchained. Great guest today that has a lot of experience in a lot of different technologies around putting data to work around the globe and how the industry has evolved over the years. I'm super excited to invite David Chapa, who is a longtime industry colleague. He's worked at competitors, he's worked as an analyst, he's worked as a friend, he's a neighbor. So it's great to have you here David.
David Chapa: Thanks a lot for having me, this is gonna be fun. I'm enjoying this already.
Molly Presley: Tell us a little bit about your background? We can see, for those that are watching on video, that you have kind of an eclectic even background behind you right now. But tell us a little bit about you being one of those unique people that's not just a technologist, but has a lot of other interests in life?
David Chapa: Yeah, so technology has always been the thing that I've been rooted in, but music is the other thing that I'm very passionate about. So I'd say there are two passions I have: technology and music. And you see a small collection of guitars behind me and actually over behind that brown guitar is Johnny Cash's road case for his Martin D35 that he used to travel with until he decided to do more air travel and he got a different case for that so it could fly in the belly of the plane.
Molly Presley: That's pretty special.
David Chapa: Yeah, that's a pretty cool thing. I don't think I'll ever part with that. You can't see it, but it does say House of Cash in the front and address and telephone number. But I've been playing guitar for almost 40 years, semi-professionally for about 25 or 30. And then I did a stint professionally for about 5, 6 years where I was doing a Johnny Cash tribute show. And I play with Johnny Cash's drummer, W.S Holland, and we had performed that show for thousands of people, and it's been a blast.
Molly Presley: You didn't do a tribute show at the Warrior Dashing Copper once, did you?
David Chapa: I did not, no.
Molly Presley: So this is embarrassing, I shouldn't even admit this to you, much less everybody who listens to the podcast, but we were up in Copper doing one of those mud runs where you do obstacles and run around the ski resort in the summer. And the night before they had a Johnny Cash tribute band, and I didn't realize it was not possible for Johnny to be playing, I thought it was actually him. And maybe there were a few cocktails involved and I was like oh my God, and I put on Facebook “I'm listening to Johnny Cash,” which I was quickly shamed. And didn't know that's not actually Johnny Cash. How funny it would've been if that had been you.
David Chapa: Yeah, that would've been funny, but that's my passion, music and technology. Matter of fact, I'm actually playing a gig this Sunday at a place locally from one to five.
Molly Presley: Nice, in Castle Rock, Colorado?
David Chapa: In Castle Rock Colorado, yeah it's fun. And what's brilliant about, I think, both music and technology, there really is a connection there because you can't see it to my right, but I've got my recording mixer over here, which is a hundred percent technology. And you think about how data flows, signal flows, this signal is the data that goes into my board, that translates into something meaningful and compelling that I can create and then deliver. And the collaboration that I do with a friend of mine in Tennessee who has the same board is incredible. So I can record music here, I can send him all the stem files, he can bring it in, and then he brings in studio musicians to add their layered parts onto the song that we were recording, the song that I wrote. So it's interesting for me as I look at what I do professionally on the technology side and the music side, and there's a lot of intersection.
Molly Presley: Absolutely, I think that's kind of a cool concept. So you can create music with somebody who's not local to you, blend it, mix it, update it, whatever you wanna do sitting from your home offices. Not unlike us podcasting from our home offices or collaborating on data science.
David Chapa: Exactly, and I remember 20 years ago working for another company, Sony was a big customer and it was a big deal for them to replicate the tracks that they were using from LA to New York for background vocals or for other things they were doing from an engineering perspective, that was a big deal. And now you look at what David Chapa can do from his home studio with a guy who has got a professional recording studio in Tennessee. I can do that very same thing. I mean the advancement that we have seen in technology, just from a hobbyist perspective is really incredible. Now take that to the next level and what we look at from the industry, and it really has opened things up considerably.
Molly Presley: Yeah, I think it is a really interesting segue into the conversation about data and how businesses are using data. You and I were talking, as we were getting ready for this conversation, about data locality and the kinds of conflicts that occur as data starts to become more localized. Maybe just let's start off the conversation there.
David Chapa: Well just take this situation here. Having all the files right here, I could very easily put it up into a shared drive: Dropbox, Sharepoint, Gdrive, which many of us do. And if you think about getting access to that from an application, there's a certain bit of latency that an application can tolerate and there's a certain bit that it can't tolerate. And so getting that data actually to a place where there is not gonna be a challenge for the application is really important. Years ago when we were talking about erasure coding, remember that?
Molly Presley: Yeah, when this new thing instead of RAIDcame out and it was done in software.
David Chapa: Yeah, and we were talking about how erasure coding would make life so much easier for people with big amounts of data because you can distribute that. You don't have to send all of it, you can distribute it. Well that's kind of where we're at, you've gotta be able to get access to the data more locally. And so we've gone beyond that, we're still using erasure coding of course, but we've gone beyond that. Getting the data as close to the application as possible allows you to be as productive as possible. Same thing with this, if I can get this data to my friend Dan as quickly as possible so that he can get access to it and get access to the musicians, their schedules, their timetables, and then get the work done, then we don't lose any money. I don't lose any money because I'm the one paying for the musicians. If it takes them longer because of the latency issues, all of a sudden that opens up Pandora's box and budgets get squashed. And in some cases, projects get either delayed or terminated.
Molly Presley: I also think that the analogy of getting a file out in let's say Google drive, the benefit you get of being able to share it, and do change control access from outside your organization. I think most of us, no matter what we do for a living, can understand the big benefit that was versus emailing around a word doc that now you had change control you weren't able to collaborate on. I can't even fathom going back to doing my job where you had to email word doc, wait for the person to give you input, email it back and then send it to the next person, I can't imagine my personal productivity not having those tools. And when you think about massive data sets, hundreds of thousands of files, many petabytes of data, whatever it is, that same concept is fantastic. If you could accomplish it with these massive data sets, but tools like SharePoint and Google Drive don't work for these massive data sets for a variety of reasons. So you think okay, why not? It's the scale, it's the latency, it's the performance, it's the ability to derive metadata and information from those files that is pretty specialized beyond what people do with their home office tools.
David Chapa: That's very true. I dunno how many years ago this was, but I know I was working for ADIC.
Molly Presley: That’s a name I haven't heard in a while, but amazing company as time.
David Chapa: Yeah, well in 2003 when I started with them, I was looking at data. I was a big backup, I still want big backup guy, backup recovery, disaster recovery, but I was looking at data and looking at the technology they were developing at the time, which was their path light VX. It was their virtual tape library. And one of the things that I said in an engineering meeting, I said at some point the metadata is going to be more valuable than the data itself. And I got a lot of pushback from everybody, but the reality is, the data around the data really is the most important thing. Now we have constructs today to protect data: data protection, backup recovery, some would say those are table stakes today. Everyone, you can do that, but getting access and understanding and discovering the data around the data, not everyone does that as well as potentially others. And so I think having that access and having that understanding is super important. That's about locality, it’s about creation, it's about access time, and it's about so many things that's critical to that data itself.
Molly Presley: You know, that brings up all these memories of things we've discussed in this industry over the years. And it makes me think a little bit about that same analogy of data versus metadata. Why does it matter? And we spent so much time figuring out how to just store data, make bigger disc drives, make higher capacity tape, make sure it's reliable, make sure you can recover, and that's all about storing the data. And back when I was working more in the backup industry, there were a lot of conversations about but that's really just insurance policies, that's your homeowner's insurance just in case. You know, we put so much effort into just the storing of data versus the using of data. And I think that's really what you're getting to around this concept of metadata is okay. We’ve, as an industry, focused for so long on just how to find a place to put all of this data that's being created so fast. And now there's a huge focus on wow, we're a data-driven world, how can we actually get value from the data insights, solve COVID, whatever it is. And that's really all about metadata and harvesting information from those big data stores. Is that generally what you were thinking back even when you were at ADIC?
David Chapa: Oh gosh, even before that, and I know the year, 1998. I was working for a large telecommunications company in the Northwest suburbs of Chicago, and it was all about the data itself. They wanted to get access to the data and run some test validation tests, whatever, they were running clear cases for development, so checking in code, checking out code, they wanted to run some tests against it for other things. And as the backup guy, I said hey, go right ahead, but if you mess up production, this is how long it's gonna take us to recover. And so at that point, I thought man, and I started using this phrase, “we are dead ending data to tape.” Now I'm not someone that's gonna be a Debbie Downer on tape, I think tape still is a good solution. I think that it has its place. Tape certainly has shifted down the continuum of data protection quite a bit, but that's where we were before, we were dead ending data to tape, and it was in a tarball. It was in a format that no one could read except the application. What was the purpose of that, other than just protecting us against failure or file deletion, right? And there was nothing else there, so if they wanted to test, I had to restore all that anyway. So that got me thinking, there's gotta be a better way.
So it was around 2000 that I started to look at just disk, give me some JBods and let me backup to disc. And so while it was still in a tarball format, it was faster to recover that because it's coming off of disc. I don't have to worry about shoe shining or anything like that with the tape. And then it was around 2002/2003 that we started to see de-duplication come in. So now we had this infinitely deep, well not infinitely, still had its limits.
Molly Presley: 10 or 20 or 40 XD program depending on what you were getting for your compression?
David Chapa: Yeah, whatever that ended up being. But then we had the challenge of recovery again with that. So we've been in a state where it's been very difficult, and fast forward, now we've got the ability to recover data into a sandbox or a playground area where you can run these types of tests and you can run these types of validations and audits very quickly. And so yeah, that's kind of where my head was going, is you've got data, let's not dead end it, let's figure out a way you can use it. It's an investment, it's a coin. Think of it as a coin, how are you gonna invest that coin?
Molly Presley: Exactly. You know, during that same timeframe, I was tracking on a similar path and this was when we created the active archive alliance and it was around the idea of okay, you've got these massive data sets largely at the time on tape libraries in a tarball, Vtar, whatever it is, some kind of proprietary format. And the idea of putting a file system in front of that, so you could use NFS, SMB standard interfaces was unlocking all kinds of opportunities for now application owners to be able to interact directly with the data. Maybe it was a bit slower than if it was on flash or disc, but you could actually interact with it, access it directly and not have to go through an IT process of recovering it out of a tarball to a file system, or to a database, whatever it is to be able to work with. And that was a big advancement being able to do that. And some of those technologies had been around a lot longer, but most general organizations, you know, HPC maybe knew about it, but the rest of the industry did not. And one of the evolutions was starting to make data accessible to the application or to the user directly. But you think about all these tools and that kind of moves on to this concept of data gravity, and because of the infrastructure-centric approach on just struggling to be able to retain as fast as it's being created, we've created these pools of data that have an enormous amount of data gravity. And you've been a bit of a pundit in addition to a technologist and you get asked these questions a lot, but what are your thoughts on data gravity? Is this something that vendors created selfishly or vendors created because there was no other answer? Is it just a lot of physics we can't overcome?
David Chapa: Yeah, so some people may not like what I'm about to say about data gravity, but I think it was a term that was developed so that we had something else to say about data. When I think of data, data carries weight when you need it. So when you're not accessing that data, it can be somewhere else and you can talk about data gravity and how you have all this data you're managing, but really, data itself is only that important to you when you actually have to have access to it. So I think about data being spread across multiple silos of storage, right? And so sometimes you do that in a hierarchical process where you're moving it to different tiers of storage. Other times that's done because IT has brought in a new tier of storage that they want to use and play with. Maybe it's faster, it's bigger, it's deeper, whatever that is, but I have a hard time with data gravity. I remember the person that first came out with that. I kind of liked it at first, and then I started to see that there was a lot of misuse of that term from the original intent. Bottom line is that for me, the data carries weight.
I think of that Dave Clark Five song He ain't heavy, he's my brother, it ain't heavy, it's my data. And my data needs to be there when you need to have access to it. So I've gotta have a system that's going to allow you to find it, get it, use it and then produce with it. Whatever that looks like, you wanna call that data gravity, let's call it data gravity. You wanna call it accessibility, let's call it data accessibility, whatever that is, but that's what it comes down to, and I brought up the word silo. I don't think anyone in IT has ever woken up in the morning and said, you know what, I really want to have four or five different silos of storage. I think that would be a really fun thing to do in my environment. And that's why the metadata is so important because when you do have these silos, you do need to understand the locality. Where is my data? What is the level of importance or that weight if you wanna give it that? And then what are my best methods and approaches for allowing my customers, my end users to have access to this data?
Molly Presley: So I think that the concept of metadata and data becomes even more important if you follow this line of thinking that generally metadata is what, a fraction of a percent of the size of the actual data set? Depending on what kind of data it is, but it's small, and it's light in comparison, right? And if you can think about moving metadata around, that's realistic, especially with 5g networks and the tools that are out there to land acceleration, things like that as you're moving around. I think that's another piece that people often don't think about. They think about “There's no way I can move a petabyte,” but there is a way you can move a fraction of a percent of a petabyte, especially if it doesn't have to be done in real-time. And that as an industry, I think as people think about I need my data silo because it has certain attributes, it has a certain cost profile, performance profile, great. IT can do that or whatever fits within budget, but if we can raise this metadata up and make it accessible and share it in other places, you can overcome this concept that you have to be local to your data because you can move your metadata around. Does that track with how you would think about it, and how the industry is evolving?
David Chapa: Exactly. And I've written about this, I published a book in 2003, I think it was the same year I started with ADIC called implementing backup and recovery. And it was about understanding the nature of the data. When you talk about disaster recovery, business continuity, one of the things I've been preaching for 30 some years is you gotta really understand the data that you're protecting so that you can restore the data that you need when you hit that wall. And that wall is a declaration of disaster. So when that happens, you need to understand what you're recovering. If you're recovering meaningless data, meaningless being maybe it's MP3s that are non-essential to run your business operation, that's taking up very valuable time. So if you can at least start with the metadata and understand the nature of the data and then create some baselines, then all of a sudden you can have a better understanding and insights into what you are managing, protecting, and potentially having to recover. So I think for me, that's where I see metadata being super valuable, is just in categorizing your data, putting them into certain buckets. You know, back in the day, I used to tell people all the time, when you get a new file system and a new storage repository, create functional directories and then put the data for those functional areas into those directories so that when you do a recovery and you set up your backup policies, you know exactly what is the priority. And of course nobody ever did that.
Molly Presley: It's very hard to control user behavior at scale, isn't it? When you think about the importance of this, and while we talk about it a lot in the data industry, I don't know if everyone really puts the pieces together because they don't sit and think about it on their drive to work or on their bike rides like we do. But this data relevant, and you think about, oh I don't know the latest data, but there's more data created every two years now than was created in the entire history of the world or whatever the latest numbers are. If you think about the evolution of things like artificial intelligence and machine learning and these applications which can derive all kinds of interesting insights that we, in our human minds, might not even think to ask or certainly couldn't process that much data. It becomes very relevant to think about okay, I spent the last 20 years of my career or my IT team spent the last 30 years storing all this data. Imagine if you could make that data accessible? And no matter if it's in a tape library or in a VTL or in a Cloud S3 bucket, whatever, to AI and ML to start to learn from or gather information, all of that data now can be used. And it's not stuck in oh gosh, I don't know where Max put it in his folder 20 years ago, so I don't know what to do with it. We can actually use all of that data, and to me that's where this conversation becomes so relevant that not only is it a forward-looking conversation, but you could actually get a lot more information from all that work you did over the last 20 years.
David Chapa: Right. So I know that data point you're talking about by the way because I think it was around 2006 that IBM research came out with this white paper on the amount of data that we are creating as a society. This was 2006, they said we are creating 2.1 quintillion bytes of data every day as a society. That's 2.1 exabytes of data, and that's 2006. If you look at IDC, look at some of the other major players in the research, they're projecting by 2025 I think it is, that the data universe should be around 184/185 zettabytes of data. Now does all that data need to be managed, protected and curated from the enterprise? No, but the fact that that data is growing at that rate. And I tell people all the time, when I get presentations, there's always a slide in there for marketing that talks about data growth. And I asked for a show of hands, how many of you know that data is growing? Okay great, everyone knows that, now let's talk about what the depth of that growth is, and how is that affecting your day-to-day? That's really what that slide is about. It's not about the fact that it's growing, it's how it is affecting you, IT managers, IT admins, and IT directors. And the fact is it's a challenge for them because to your point, how do I find Max's golden chest of treasures that he created when he was here and now he's no longer here for us to contact to find out. That's the problem I think we're seeing with a lot of management over the data in these silos.
Molly Presley: So one of the areas that I've had a really interesting executive staff conversations on in the last five or so years has been as data scientists have come into the organization. And you have these folks who are coming outta school largely that are code horses. They are analytical, they're ready to do data science, and then they show up as a team and say okay, I need my data IT team. And in some cases, there's a data engineer creating the data purpose built for what they're doing, but in a lot of cases, they need to harvest those old systems. And so IT gets put in an even tougher spot that now not only are they trying to serve their budgets and their data growth, but they're also trying to feed this new team of data scientists who can't self-serve really because of the way that data's been stored. And so it has really created even more pressure. I always think that all pressure ends up coming still on IT in these environments because they have to be involved because the way the architectures were set up over the last 20 years and by no fault of their own, it's what was available.
David Chapa: A hundred percent. You bring up data scientists, I feel like I need to put on my beard so I look more data scientist-like. When that whole title started to become more relevant in IT, and your chief data officer, one of the things that I asked a lot of questions about is that role and what the intent is of that role. And I got the best definition from a guy in Singapore, he said “The most important part of my role is to know what questions to ask of the data.” I'm like, that's great, I said, so you get access to that data immediately, and then you can create your data schemas. He's like, well getting access to the data is different than knowing what questions to ask of the data. And so to your point, how do I extract that value? How do I get that? And it's not just the data that's spinning on disc, it's data that's under management. And I think that we sometimes forget that data under management might still be in a tarball, and might be on an LTO cartridge.
Molly Presley: Or on a scientist machine next to a sequencer, right?
David Chapa: Right, yeah, so you've gotta be able to have access to that data. You know, as you were talking, I was thinking of one of my clients and I don't think I should mention them, but it was a financial company up in Menomonee Falls, Wisconsin, which was a great thing to say, Menomonee Falls. I was consulting with them, I was on a three month contract and I stayed there for like two years consulting with them on data protection and disaster recovery. One of the things that they did was they made duplications of their tape and sent it off site forever irrespective of expiration. I told them that was a major exposure, you needed to bring that data back or those tapes back, and you need to get rid of it. If you get subpoenaed, they could have access potentially to anything else that's not expired. The response was we have enough money, and if we get sued, then we can deal with it.
Well they came under investigation, and long story short some trading abnormalities were found. And basically the company was acquired by Wells Fargo and the executives were barred from trading for life, that type of thing. Again, having access to the metadata, knowing what data you have under management is super important. It can be the difference of knowing what to ask of your data from a data scientist’s perspective or what we legally need to maintain and retain and what we can legally get rid of. And so this is an understanding of data that goes beyond just wanting to know how many MP3 files we have on our system. This goes into much greater depths. You know, what are we protecting? Okay, we're protecting our email, great. Now how long do we have to keep that for? 31 days, great. How much of that are we keeping that's beyond 31 days? We need to eradicate that data. So the metadata goes beyond just even what you think is the normal use of metadata in categorizing, understanding, locality, creation, but also the after management of it.
.
Molly Presley: Absolutely, at Hammerspace, the company who I work for, we're using a term which isn't new in the industry, but it hasn't been used that much around data objectives. And so instead of setting a policy and trying to remember my various compliance rules, delete this after this amount of time or archive this after that amount of time. Being able to set objectives as a business or as an IT administrator saying my objective is I want to protect against this, meet this compliance rule, and then have the system just automatically take care of it across silos and softwares and things like that. But those objectives have been difficult to translate into a myriad of policies and different software and storage systems just because there's so much to do. And that's another challenge that IT always ends up in, but that's really what it's about, you have to meet your business objectives, and sometimes that's innovation, sometimes that's protection. Sometimes it's avoiding ransomware, and the more we can do as vendors to help make that easier and maybe more automated, the more successful organizations will be in meeting their goals.
Thinking about vendors just for a minute before we tie up, we've talked a lot about vendor technologies, proprietary things like proprietary tarballs, things like that, where do you think our industry should go in the next 10 years to help companies accomplish their objectives? And I'm thinking about open source, I'm thinking about ease of accessibility, any of those things, what would just jump to the top of your mind as far as what vendors can do to make this easier on organizations?
David Chapa: Boy, that's a loaded question. There are a lot of things that I think that we could do, and I think it's a challenge for some organizations to say you know what, I'm gonna turn the key on this and make it open source. But there are certain areas where I think that's important, that's a critical thing to do for the greater good of IT. And in some cases, national security, you talk about ransomware. I think from the ransomware side, which is something that I've been looking at quite a bit, there's gotta be something more that's happening. I don't have the accurate stats here, but I'm pretty confident that the point in time when customers identify that they have been infected with malware, ransomware, what have you, is it's not when they get the ransomware, it's usually about 80 days in the past when they first were affected.
Molly Presley: Interesting.
David Chapa: And years ago, like in 2015, that number was like 222 days. So understanding the data movement, understanding anomalies, I think that's the thing that we need to look at really in managing data, are those anomalies. Creating a baseline, and then from that baseline, if I see anything that goes off, up and down, it's like a heartbeat. If your resting heart rate is X and I see that it goes up by 7 or 10 points, there's something alarming there. I wear a fitness band that tracks my heart rate, and if it goes up above a certain heart rate, it's gonna say either you're working really hard, or you need to call 911. So those types of things, I think that we need to look at from a dashboard perspective. It's not binary, doesn't say okay, we are infected, we couldn't in fact have a migration going on, so we have multiple terabytes of data being moved.
Molly Presley: Sure, but alert the human to take a look.
David Chapa: That's my point, right. I've found an anomaly, I am your security bot or whatever that is, and you send that off to the SIM. And now your network security operations guys and gals are looking at that, trying to identify okay, what's really happening here? I think that's one of the areas that we should be looking at bolstering quite a bit.
Molly Presley: That makes a lot of sense. I mean, you think about everything we're talking about is really not manageable by a human being, no matter how intelligent the human being or team of human beings is that automation where software can. They're never gonna replace humans or at least not in the foreseeable future. But having software helps them know where to take a look, where to automate things, instead of having to manually build policies or write code script things. A lot of things make humans more efficient. I would tend to agree that that would help the industry at large and something vendors really could help to do.
David Chapa: Right, exactly. And that's what I think. I think on another side, we’re not necessarily open source, but we've been talking about this since 2015. The department of energy has been issuing grants to different organizations to build the first exaflop supercomputer, and we're not there yet. And I think one of the reasons we're not there yet is because we haven't found a file system yet that can actually feed the cores of an exaflop supercomputer. But I think once we reach that level, I think from a research perspective, that's gonna change the game for us not only nationally here in the United States, but globally for everybody. I mean it's gonna be a good thing for us to reach that level. And it was supposed to be delivered in 2020, we all know what happened in 2020, and I think that 2025 is the new date that we're looking at, but I'll be excited when we actually see a true exaflop supercomputer. That means that we've also developed a file system that can drive those cores which is gonna be very important, but absolutely.
Molly Presley: Absolutely, I could talk to you about all kinds of different topics on this, and it is interesting that idea of scale. So exaflop requires massive performance, read-write, IO patterns, but also massive scale. And this is an interesting thing that object storage doesn't overcome. So object storage is massive in the number of objects, but it doesn't serve that purpose of an exaflop file system either. So it's interesting that as you watch how innovation has occurred, it's been kind of like okay, let's pull the operating system out of the server and put it in the storage array. Okay, now let's scale across arrays and then it was like okay great, build object storage, now we're back to but we still need that file system and we need it at scale. And I know there are a lot of very interesting conversations occurring there on the innovation side. But it's a great way to tie it to some of these national security as well as it helps drive our economy. The countries with the biggest supercomputers or the most compute power tend to be the economic leads too. It's interesting that you can almost track it directly. When you look at the top 10 largest supercomputers, those are usually the largest economies too, so it all matters in interesting interconnected ways.
David Chapa: I think the fastest supercomputer still to date is in Japan, and I think it's at 415 petaflops.
Molly Presley:I haven't looked at the list in the last couple of years, but it certainly could
be.
David Chapa: I looked within the last six months, I'm pretty sure it's still in Japan. Now there are some strides being made, like Invidia with their GPU technology and they're working with AMD, I believe it's AMD. They've been able to get bursts at exaflop, but it's not as consistent. So I'm hopeful as I see this stuff happen. And I know this is completely off topic here, but here's the cool thing about this is when I start looking at the supercomputer end of things, it reminds me a lot of the Unix community where everyone is willing to share anything and everything you have to try to reach the end goal. It's kinda like open source on steroids; the supercomputer world.
Molly Presley: Yeah, it totally is.
David Chapa: And so reading this stuff has been just fascinating for me. And seeing what people are doing and learning about what file systems and open source potentially could reach those levels. And then what file systems I'm looking at in the market today that I think also could hit those levels probably more quickly.
Molly Presley: Yep, absolutely. Well David, this has been great having you on. I tried to throw you a curve ball at the beginning and get you to play some music for us, and I didn't give you enough notice. But thinking of this collaborative distributed world we live in, where can we hear some of your music? I certainly know where we can find a lot of your technology insights.
David Chapa: So Spotify, they put up an EP of some original songs off an album that I recorded called the Storyteller, and so that's on Spotify. That’s an album, actually the EP has W.S Holland playing drums. Now W.S Holland was Johnny Cash's drummer for almost 40 years, and he and I played together for five years until he passed away in 2020, but the last two albums he ever recorded were mine. So I feel like that's a pretty good thing to have on my musical resume, but you could also Google David Chapa Music and what you may find are my YouTube channel, and you’ll also find some spots on the Fox morning show in Branson, Missouri where I'd always stop off before heading to Tennessee to do some recording. So David Chapa Music on Facebook, Instagram, and YouTube.
Molly Presley: Cool, thanks so much for joining today David, always fun to catch up with you and I'm sure we'll chat again.
David Chapa: I hope so, thank you.
Molly Presley: Hello, and welcome to this episode of Data Unchained. Great guest today that has a lot of experience in a lot of different technologies around putting data to work around the globe and how the industry has evolved over the years. I'm super excited to invite David Chapa, who is a longtime industry colleague. He's worked at competitors, he's worked as an analyst, he's worked as a friend, he's a neighbor. So it's great to have you here David.
David Chapa: Thanks a lot for having me, this is gonna be fun. I'm enjoying this already.
Molly Presley: Tell us a little bit about your background? We can see, for those that are watching on video, that you have kind of an eclectic even background behind you right now. But tell us a little bit about you being one of those unique people that's not just a technologist, but has a lot of other interests in life?
David Chapa: Yeah, so technology has always been the thing that I've been rooted in, but music is the other thing that I'm very passionate about. So I'd say there are two passions I have: technology and music. And you see a small collection of guitars behind me and actually over behind that brown guitar is Johnny Cash's road case for his Martin D35 that he used to travel with until he decided to do more air travel and he got a different case for that so it could fly in the belly of the plane.
Molly Presley: That's pretty special.
David Chapa: Yeah, that's a pretty cool thing. I don't think I'll ever part with that. You can't see it, but it does say House of Cash in the front and address and telephone number. But I've been playing guitar for almost 40 years, semi-professionally for about 25 or 30. And then I did a stint professionally for about 5, 6 years where I was doing a Johnny Cash tribute show. And I play with Johnny Cash's drummer, W.S Holland, and we had performed that show for thousands of people, and it's been a blast.
Molly Presley: You didn't do a tribute show at the Warrior Dashing Copper once, did you?
David Chapa: I did not, no.
Molly Presley: So this is embarrassing, I shouldn't even admit this to you, much less everybody who listens to the podcast, but we were up in Copper doing one of those mud runs where you do obstacles and run around the ski resort in the summer. And the night before they had a Johnny Cash tribute band, and I didn't realize it was not possible for Johnny to be playing, I thought it was actually him. And maybe there were a few cocktails involved and I was like oh my God, and I put on Facebook “I'm listening to Johnny Cash,” which I was quickly shamed. And didn't know that's not actually Johnny Cash. How funny it would've been if that had been you.
David Chapa: Yeah, that would've been funny, but that's my passion, music and technology. Matter of fact, I'm actually playing a gig this Sunday at a place locally from one to five.
Molly Presley: Nice, in Castle Rock, Colorado?
David Chapa: In Castle Rock Colorado, yeah it's fun. And what's brilliant about, I think, both music and technology, there really is a connection there because you can't see it to my right, but I've got my recording mixer over here, which is a hundred percent technology. And you think about how data flows, signal flows, this signal is the data that goes into my board, that translates into something meaningful and compelling that I can create and then deliver. And the collaboration that I do with a friend of mine in Tennessee who has the same board is incredible. So I can record music here, I can send him all the stem files, he can bring it in, and then he brings in studio musicians to add their layered parts onto the song that we were recording, the song that I wrote. So it's interesting for me as I look at what I do professionally on the technology side and the music side, and there's a lot of intersection.
Molly Presley: Absolutely, I think that's kind of a cool concept. So you can create music with somebody who's not local to you, blend it, mix it, update it, whatever you wanna do sitting from your home offices. Not unlike us podcasting from our home offices or collaborating on data science.
David Chapa: Exactly, and I remember 20 years ago working for another company, Sony was a big customer and it was a big deal for them to replicate the tracks that they were using from LA to New York for background vocals or for other things they were doing from an engineering perspective, that was a big deal. And now you look at what David Chapa can do from his home studio with a guy who has got a professional recording studio in Tennessee. I can do that very same thing. I mean the advancement that we have seen in technology, just from a hobbyist perspective is really incredible. Now take that to the next level and what we look at from the industry, and it really has opened things up considerably.
Molly Presley: Yeah, I think it is a really interesting segue into the conversation about data and how businesses are using data. You and I were talking, as we were getting ready for this conversation, about data locality and the kinds of conflicts that occur as data starts to become more localized. Maybe just let's start off the conversation there.
David Chapa: Well just take this situation here. Having all the files right here, I could very easily put it up into a shared drive: Dropbox, Sharepoint, Gdrive, which many of us do. And if you think about getting access to that from an application, there's a certain bit of latency that an application can tolerate and there's a certain bit that it can't tolerate. And so getting that data actually to a place where there is not gonna be a challenge for the application is really important. Years ago when we were talking about erasure coding, remember that?
Molly Presley: Yeah, when this new thing instead of RAIDcame out and it was done in software.
David Chapa: Yeah, and we were talking about how erasure coding would make life so much easier for people with big amounts of data because you can distribute that. You don't have to send all of it, you can distribute it. Well that's kind of where we're at, you've gotta be able to get access to the data more locally. And so we've gone beyond that, we're still using erasure coding of course, but we've gone beyond that. Getting the data as close to the application as possible allows you to be as productive as possible. Same thing with this, if I can get this data to my friend Dan as quickly as possible so that he can get access to it and get access to the musicians, their schedules, their timetables, and then get the work done, then we don't lose any money. I don't lose any money because I'm the one paying for the musicians. If it takes them longer because of the latency issues, all of a sudden that opens up Pandora's box and budgets get squashed. And in some cases, projects get either delayed or terminated.
Molly Presley: I also think that the analogy of getting a file out in let's say Google drive, the benefit you get of being able to share it, and do change control access from outside your organization. I think most of us, no matter what we do for a living, can understand the big benefit that was versus emailing around a word doc that now you had change control you weren't able to collaborate on. I can't even fathom going back to doing my job where you had to email word doc, wait for the person to give you input, email it back and then send it to the next person, I can't imagine my personal productivity not having those tools. And when you think about massive data sets, hundreds of thousands of files, many petabytes of data, whatever it is, that same concept is fantastic. If you could accomplish it with these massive data sets, but tools like SharePoint and Google Drive don't work for these massive data sets for a variety of reasons. So you think okay, why not? It's the scale, it's the latency, it's the performance, it's the ability to derive metadata and information from those files that is pretty specialized beyond what people do with their home office tools.
David Chapa: That's very true. I dunno how many years ago this was, but I know I was working for ADIC.
Molly Presley: That’s a name I haven't heard in a while, but amazing company as time.
David Chapa: Yeah, well in 2003 when I started with them, I was looking at data. I was a big backup, I still want big backup guy, backup recovery, disaster recovery, but I was looking at data and looking at the technology they were developing at the time, which was their path light VX. It was their virtual tape library. And one of the things that I said in an engineering meeting, I said at some point the metadata is going to be more valuable than the data itself. And I got a lot of pushback from everybody, but the reality is, the data around the data really is the most important thing. Now we have constructs today to protect data: data protection, backup recovery, some would say those are table stakes today. Everyone, you can do that, but getting access and understanding and discovering the data around the data, not everyone does that as well as potentially others. And so I think having that access and having that understanding is super important. That's about locality, it’s about creation, it's about access time, and it's about so many things that's critical to that data itself.
Molly Presley: You know, that brings up all these memories of things we've discussed in this industry over the years. And it makes me think a little bit about that same analogy of data versus metadata. Why does it matter? And we spent so much time figuring out how to just store data, make bigger disc drives, make higher capacity tape, make sure it's reliable, make sure you can recover, and that's all about storing the data. And back when I was working more in the backup industry, there were a lot of conversations about but that's really just insurance policies, that's your homeowner's insurance just in case. You know, we put so much effort into just the storing of data versus the using of data. And I think that's really what you're getting to around this concept of metadata is okay. We’ve, as an industry, focused for so long on just how to find a place to put all of this data that's being created so fast. And now there's a huge focus on wow, we're a data-driven world, how can we actually get value from the data insights, solve COVID, whatever it is. And that's really all about metadata and harvesting information from those big data stores. Is that generally what you were thinking back even when you were at ADIC?
David Chapa: Oh gosh, even before that, and I know the year, 1998. I was working for a large telecommunications company in the Northwest suburbs of Chicago, and it was all about the data itself. They wanted to get access to the data and run some test validation tests, whatever, they were running clear cases for development, so checking in code, checking out code, they wanted to run some tests against it for other things. And as the backup guy, I said hey, go right ahead, but if you mess up production, this is how long it's gonna take us to recover. And so at that point, I thought man, and I started using this phrase, “we are dead ending data to tape.” Now I'm not someone that's gonna be a Debbie Downer on tape, I think tape still is a good solution. I think that it has its place. Tape certainly has shifted down the continuum of data protection quite a bit, but that's where we were before, we were dead ending data to tape, and it was in a tarball. It was in a format that no one could read except the application. What was the purpose of that, other than just protecting us against failure or file deletion, right? And there was nothing else there, so if they wanted to test, I had to restore all that anyway. So that got me thinking, there's gotta be a better way.
So it was around 2000 that I started to look at just disk, give me some JBods and let me backup to disc. And so while it was still in a tarball format, it was faster to recover that because it's coming off of disc. I don't have to worry about shoe shining or anything like that with the tape. And then it was around 2002/2003 that we started to see de-duplication come in. So now we had this infinitely deep, well not infinitely, still had its limits.
Molly Presley: 10 or 20 or 40 XD program depending on what you were getting for your compression?
David Chapa: Yeah, whatever that ended up being. But then we had the challenge of recovery again with that. So we've been in a state where it's been very difficult, and fast forward, now we've got the ability to recover data into a sandbox or a playground area where you can run these types of tests and you can run these types of validations and audits very quickly. And so yeah, that's kind of where my head was going, is you've got data, let's not dead end it, let's figure out a way you can use it. It's an investment, it's a coin. Think of it as a coin, how are you gonna invest that coin?
Molly Presley: Exactly. You know, during that same timeframe, I was tracking on a similar path and this was when we created the active archive alliance and it was around the idea of okay, you've got these massive data sets largely at the time on tape libraries in a tarball, Vtar, whatever it is, some kind of proprietary format. And the idea of putting a file system in front of that, so you could use NFS, SMB standard interfaces was unlocking all kinds of opportunities for now application owners to be able to interact directly with the data. Maybe it was a bit slower than if it was on flash or disc, but you could actually interact with it, access it directly and not have to go through an IT process of recovering it out of a tarball to a file system, or to a database, whatever it is to be able to work with. And that was a big advancement being able to do that. And some of those technologies had been around a lot longer, but most general organizations, you know, HPC maybe knew about it, but the rest of the industry did not. And one of the evolutions was starting to make data accessible to the application or to the user directly. But you think about all these tools and that kind of moves on to this concept of data gravity, and because of the infrastructure-centric approach on just struggling to be able to retain as fast as it's being created, we've created these pools of data that have an enormous amount of data gravity. And you've been a bit of a pundit in addition to a technologist and you get asked these questions a lot, but what are your thoughts on data gravity? Is this something that vendors created selfishly or vendors created because there was no other answer? Is it just a lot of physics we can't overcome?
David Chapa: Yeah, so some people may not like what I'm about to say about data gravity, but I think it was a term that was developed so that we had something else to say about data. When I think of data, data carries weight when you need it. So when you're not accessing that data, it can be somewhere else and you can talk about data gravity and how you have all this data you're managing, but really, data itself is only that important to you when you actually have to have access to it. So I think about data being spread across multiple silos of storage, right? And so sometimes you do that in a hierarchical process where you're moving it to different tiers of storage. Other times that's done because IT has brought in a new tier of storage that they want to use and play with. Maybe it's faster, it's bigger, it's deeper, whatever that is, but I have a hard time with data gravity. I remember the person that first came out with that. I kind of liked it at first, and then I started to see that there was a lot of misuse of that term from the original intent. Bottom line is that for me, the data carries weight.
I think of that Dave Clark Five song He ain't heavy, he's my brother, it ain't heavy, it's my data. And my data needs to be there when you need to have access to it. So I've gotta have a system that's going to allow you to find it, get it, use it and then produce with it. Whatever that looks like, you wanna call that data gravity, let's call it data gravity. You wanna call it accessibility, let's call it data accessibility, whatever that is, but that's what it comes down to, and I brought up the word silo. I don't think anyone in IT has ever woken up in the morning and said, you know what, I really want to have four or five different silos of storage. I think that would be a really fun thing to do in my environment. And that's why the metadata is so important because when you do have these silos, you do need to understand the locality. Where is my data? What is the level of importance or that weight if you wanna give it that? And then what are my best methods and approaches for allowing my customers, my end users to have access to this data?
Molly Presley: So I think that the concept of metadata and data becomes even more important if you follow this line of thinking that generally metadata is what, a fraction of a percent of the size of the actual data set? Depending on what kind of data it is, but it's small, and it's light in comparison, right? And if you can think about moving metadata around, that's realistic, especially with 5g networks and the tools that are out there to land acceleration, things like that as you're moving around. I think that's another piece that people often don't think about. They think about “There's no way I can move a petabyte,” but there is a way you can move a fraction of a percent of a petabyte, especially if it doesn't have to be done in real-time. And that as an industry, I think as people think about I need my data silo because it has certain attributes, it has a certain cost profile, performance profile, great. IT can do that or whatever fits within budget, but if we can raise this metadata up and make it accessible and share it in other places, you can overcome this concept that you have to be local to your data because you can move your metadata around. Does that track with how you would think about it, and how the industry is evolving?
David Chapa: Exactly. And I've written about this, I published a book in 2003, I think it was the same year I started with ADIC called implementing backup and recovery. And it was about understanding the nature of the data. When you talk about disaster recovery, business continuity, one of the things I've been preaching for 30 some years is you gotta really understand the data that you're protecting so that you can restore the data that you need when you hit that wall. And that wall is a declaration of disaster. So when that happens, you need to understand what you're recovering. If you're recovering meaningless data, meaningless being maybe it's MP3s that are non-essential to run your business operation, that's taking up very valuable time. So if you can at least start with the metadata and understand the nature of the data and then create some baselines, then all of a sudden you can have a better understanding and insights into what you are managing, protecting, and potentially having to recover. So I think for me, that's where I see metadata being super valuable, is just in categorizing your data, putting them into certain buckets. You know, back in the day, I used to tell people all the time, when you get a new file system and a new storage repository, create functional directories and then put the data for those functional areas into those directories so that when you do a recovery and you set up your backup policies, you know exactly what is the priority. And of course nobody ever did that.
Molly Presley: It's very hard to control user behavior at scale, isn't it? When you think about the importance of this, and while we talk about it a lot in the data industry, I don't know if everyone really puts the pieces together because they don't sit and think about it on their drive to work or on their bike rides like we do. But this data relevant, and you think about, oh I don't know the latest data, but there's more data created every two years now than was created in the entire history of the world or whatever the latest numbers are. If you think about the evolution of things like artificial intelligence and machine learning and these applications which can derive all kinds of interesting insights that we, in our human minds, might not even think to ask or certainly couldn't process that much data. It becomes very relevant to think about okay, I spent the last 20 years of my career or my IT team spent the last 30 years storing all this data. Imagine if you could make that data accessible? And no matter if it's in a tape library or in a VTL or in a Cloud S3 bucket, whatever, to AI and ML to start to learn from or gather information, all of that data now can be used. And it's not stuck in oh gosh, I don't know where Max put it in his folder 20 years ago, so I don't know what to do with it. We can actually use all of that data, and to me that's where this conversation becomes so relevant that not only is it a forward-looking conversation, but you could actually get a lot more information from all that work you did over the last 20 years.
David Chapa: Right. So I know that data point you're talking about by the way because I think it was around 2006 that IBM research came out with this white paper on the amount of data that we are creating as a society. This was 2006, they said we are creating 2.1 quintillion bytes of data every day as a society. That's 2.1 exabytes of data, and that's 2006. If you look at IDC, look at some of the other major players in the research, they're projecting by 2025 I think it is, that the data universe should be around 184/185 zettabytes of data. Now does all that data need to be managed, protected and curated from the enterprise? No, but the fact that that data is growing at that rate. And I tell people all the time, when I get presentations, there's always a slide in there for marketing that talks about data growth. And I asked for a show of hands, how many of you know that data is growing? Okay great, everyone knows that, now let's talk about what the depth of that growth is, and how is that affecting your day-to-day? That's really what that slide is about. It's not about the fact that it's growing, it's how it is affecting you, IT managers, IT admins, and IT directors. And the fact is it's a challenge for them because to your point, how do I find Max's golden chest of treasures that he created when he was here and now he's no longer here for us to contact to find out. That's the problem I think we're seeing with a lot of management over the data in these silos.
Molly Presley: So one of the areas that I've had a really interesting executive staff conversations on in the last five or so years has been as data scientists have come into the organization. And you have these folks who are coming outta school largely that are code horses. They are analytical, they're ready to do data science, and then they show up as a team and say okay, I need my data IT team. And in some cases, there's a data engineer creating the data purpose built for what they're doing, but in a lot of cases, they need to harvest those old systems. And so IT gets put in an even tougher spot that now not only are they trying to serve their budgets and their data growth, but they're also trying to feed this new team of data scientists who can't self-serve really because of the way that data's been stored. And so it has really created even more pressure. I always think that all pressure ends up coming still on IT in these environments because they have to be involved because the way the architectures were set up over the last 20 years and by no fault of their own, it's what was available.
David Chapa: A hundred percent. You bring up data scientists, I feel like I need to put on my beard so I look more data scientist-like. When that whole title started to become more relevant in IT, and your chief data officer, one of the things that I asked a lot of questions about is that role and what the intent is of that role. And I got the best definition from a guy in Singapore, he said “The most important part of my role is to know what questions to ask of the data.” I'm like, that's great, I said, so you get access to that data immediately, and then you can create your data schemas. He's like, well getting access to the data is different than knowing what questions to ask of the data. And so to your point, how do I extract that value? How do I get that? And it's not just the data that's spinning on disc, it's data that's under management. And I think that we sometimes forget that data under management might still be in a tarball, and might be on an LTO cartridge.
Molly Presley: Or on a scientist machine next to a sequencer, right?
David Chapa: Right, yeah, so you've gotta be able to have access to that data. You know, as you were talking, I was thinking of one of my clients and I don't think I should mention them, but it was a financial company up in Menomonee Falls, Wisconsin, which was a great thing to say, Menomonee Falls. I was consulting with them, I was on a three month contract and I stayed there for like two years consulting with them on data protection and disaster recovery. One of the things that they did was they made duplications of their tape and sent it off site forever irrespective of expiration. I told them that was a major exposure, you needed to bring that data back or those tapes back, and you need to get rid of it. If you get subpoenaed, they could have access potentially to anything else that's not expired. The response was we have enough money, and if we get sued, then we can deal with it.
Well they came under investigation, and long story short some trading abnormalities were found. And basically the company was acquired by Wells Fargo and the executives were barred from trading for life, that type of thing. Again, having access to the metadata, knowing what data you have under management is super important. It can be the difference of knowing what to ask of your data from a data scientist’s perspective or what we legally need to maintain and retain and what we can legally get rid of. And so this is an understanding of data that goes beyond just wanting to know how many MP3 files we have on our system. This goes into much greater depths. You know, what are we protecting? Okay, we're protecting our email, great. Now how long do we have to keep that for? 31 days, great. How much of that are we keeping that's beyond 31 days? We need to eradicate that data. So the metadata goes beyond just even what you think is the normal use of metadata in categorizing, understanding, locality, creation, but also the after management of it.
.
Molly Presley: Absolutely, at Hammerspace, the company who I work for, we're using a term which isn't new in the industry, but it hasn't been used that much around data objectives. And so instead of setting a policy and trying to remember my various compliance rules, delete this after this amount of time or archive this after that amount of time. Being able to set objectives as a business or as an IT administrator saying my objective is I want to protect against this, meet this compliance rule, and then have the system just automatically take care of it across silos and softwares and things like that. But those objectives have been difficult to translate into a myriad of policies and different software and storage systems just because there's so much to do. And that's another challenge that IT always ends up in, but that's really what it's about, you have to meet your business objectives, and sometimes that's innovation, sometimes that's protection. Sometimes it's avoiding ransomware, and the more we can do as vendors to help make that easier and maybe more automated, the more successful organizations will be in meeting their goals.
Thinking about vendors just for a minute before we tie up, we've talked a lot about vendor technologies, proprietary things like proprietary tarballs, things like that, where do you think our industry should go in the next 10 years to help companies accomplish their objectives? And I'm thinking about open source, I'm thinking about ease of accessibility, any of those things, what would just jump to the top of your mind as far as what vendors can do to make this easier on organizations?
David Chapa: Boy, that's a loaded question. There are a lot of things that I think that we could do, and I think it's a challenge for some organizations to say you know what, I'm gonna turn the key on this and make it open source. But there are certain areas where I think that's important, that's a critical thing to do for the greater good of IT. And in some cases, national security, you talk about ransomware. I think from the ransomware side, which is something that I've been looking at quite a bit, there's gotta be something more that's happening. I don't have the accurate stats here, but I'm pretty confident that the point in time when customers identify that they have been infected with malware, ransomware, what have you, is it's not when they get the ransomware, it's usually about 80 days in the past when they first were affected.
Molly Presley: Interesting.
David Chapa: And years ago, like in 2015, that number was like 222 days. So understanding the data movement, understanding anomalies, I think that's the thing that we need to look at really in managing data, are those anomalies. Creating a baseline, and then from that baseline, if I see anything that goes off, up and down, it's like a heartbeat. If your resting heart rate is X and I see that it goes up by 7 or 10 points, there's something alarming there. I wear a fitness band that tracks my heart rate, and if it goes up above a certain heart rate, it's gonna say either you're working really hard, or you need to call 911. So those types of things, I think that we need to look at from a dashboard perspective. It's not binary, doesn't say okay, we are infected, we couldn't in fact have a migration going on, so we have multiple terabytes of data being moved.
Molly Presley: Sure, but alert the human to take a look.
David Chapa: That's my point, right. I've found an anomaly, I am your security bot or whatever that is, and you send that off to the SIM. And now your network security operations guys and gals are looking at that, trying to identify okay, what's really happening here? I think that's one of the areas that we should be looking at bolstering quite a bit.
Molly Presley: That makes a lot of sense. I mean, you think about everything we're talking about is really not manageable by a human being, no matter how intelligent the human being or team of human beings is that automation where software can. They're never gonna replace humans or at least not in the foreseeable future. But having software helps them know where to take a look, where to automate things, instead of having to manually build policies or write code script things. A lot of things make humans more efficient. I would tend to agree that that would help the industry at large and something vendors really could help to do.
David Chapa: Right, exactly. And that's what I think. I think on another side, we’re not necessarily open source, but we've been talking about this since 2015. The department of energy has been issuing grants to different organizations to build the first exaflop supercomputer, and we're not there yet. And I think one of the reasons we're not there yet is because we haven't found a file system yet that can actually feed the cores of an exaflop supercomputer. But I think once we reach that level, I think from a research perspective, that's gonna change the game for us not only nationally here in the United States, but globally for everybody. I mean it's gonna be a good thing for us to reach that level. And it was supposed to be delivered in 2020, we all know what happened in 2020, and I think that 2025 is the new date that we're looking at, but I'll be excited when we actually see a true exaflop supercomputer. That means that we've also developed a file system that can drive those cores which is gonna be very important, but absolutely.
Molly Presley: Absolutely, I could talk to you about all kinds of different topics on this, and it is interesting that idea of scale. So exaflop requires massive performance, read-write, IO patterns, but also massive scale. And this is an interesting thing that object storage doesn't overcome. So object storage is massive in the number of objects, but it doesn't serve that purpose of an exaflop file system either. So it's interesting that as you watch how innovation has occurred, it's been kind of like okay, let's pull the operating system out of the server and put it in the storage array. Okay, now let's scale across arrays and then it was like okay great, build object storage, now we're back to but we still need that file system and we need it at scale. And I know there are a lot of very interesting conversations occurring there on the innovation side. But it's a great way to tie it to some of these national security as well as it helps drive our economy. The countries with the biggest supercomputers or the most compute power tend to be the economic leads too. It's interesting that you can almost track it directly. When you look at the top 10 largest supercomputers, those are usually the largest economies too, so it all matters in interesting interconnected ways.
David Chapa: I think the fastest supercomputer still to date is in Japan, and I think it's at 415 petaflops.
Molly Presley:I haven't looked at the list in the last couple of years, but it certainly could
be.
David Chapa: I looked within the last six months, I'm pretty sure it's still in Japan. Now there are some strides being made, like Invidia with their GPU technology and they're working with AMD, I believe it's AMD. They've been able to get bursts at exaflop, but it's not as consistent. So I'm hopeful as I see this stuff happen. And I know this is completely off topic here, but here's the cool thing about this is when I start looking at the supercomputer end of things, it reminds me a lot of the Unix community where everyone is willing to share anything and everything you have to try to reach the end goal. It's kinda like open source on steroids; the supercomputer world.
Molly Presley: Yeah, it totally is.
David Chapa: And so reading this stuff has been just fascinating for me. And seeing what people are doing and learning about what file systems and open source potentially could reach those levels. And then what file systems I'm looking at in the market today that I think also could hit those levels probably more quickly.
Molly Presley: Yep, absolutely. Well David, this has been great having you on. I tried to throw you a curve ball at the beginning and get you to play some music for us, and I didn't give you enough notice. But thinking of this collaborative distributed world we live in, where can we hear some of your music? I certainly know where we can find a lot of your technology insights.
David Chapa: So Spotify, they put up an EP of some original songs off an album that I recorded called the Storyteller, and so that's on Spotify. That’s an album, actually the EP has W.S Holland playing drums. Now W.S Holland was Johnny Cash's drummer for almost 40 years, and he and I played together for five years until he passed away in 2020, but the last two albums he ever recorded were mine. So I feel like that's a pretty good thing to have on my musical resume, but you could also Google David Chapa Music and what you may find are my YouTube channel, and you’ll also find some spots on the Fox morning show in Branson, Missouri where I'd always stop off before heading to Tennessee to do some recording. So David Chapa Music on Facebook, Instagram, and YouTube.
Molly Presley: Cool, thanks so much for joining today David, always fun to catch up with you and I'm sure we'll chat again.
David Chapa: I hope so, thank you.
