1 00:00:00,000 --> 00:00:03,780 Hunter Leath: That was the goal of ours was to be in a position 2 00:00:03,780 --> 00:00:07,950 to offer this experience like EFS without having to charge the prices. 3 00:00:08,235 --> 00:00:12,015 That I know so many customers are frustrated with from a thing like EFS. 4 00:00:17,595 --> 00:00:19,305 Corey Quinn: Welcome to Screaming in the Cloud. 5 00:00:19,545 --> 00:00:20,835 I'm Cory Quinn. 6 00:00:21,134 --> 00:00:25,135 I'm joined today by Hunter Leath, the CEO over at Archil. 7 00:00:25,365 --> 00:00:29,840 Uh, I get a. Bunch of questions periodically from folks who are building 8 00:00:29,840 --> 00:00:34,130 something new in the AWS ecosystem, and they want me to take a look at it. 9 00:00:34,190 --> 00:00:37,460 And I, I generally turn most of them down because I'm a skeptic. 10 00:00:37,460 --> 00:00:41,660 But for whatever reason, hunter caught me on a good day and I looked at it and I 11 00:00:41,750 --> 00:00:44,300 really liked what it was he was building. 12 00:00:44,780 --> 00:00:47,420 Hunter, thank you for joining me and suffering my slings 13 00:00:47,420 --> 00:00:48,830 and arrows now public version. 14 00:00:49,460 --> 00:00:50,269 Hunter Leath: Thanks, Corey. 15 00:00:50,390 --> 00:00:51,320 Excited to be here. 16 00:00:52,080 --> 00:00:55,170 Corey Quinn: This episode is sponsored in part by my day job Duck. 17 00:00:55,170 --> 00:00:58,379 Bill, do you have a horrifying AWS bill? 18 00:00:58,650 --> 00:01:00,540 That can mean a lot of things. 19 00:01:00,750 --> 00:01:04,739 Predicting what it's going to be, determining what it should be, 20 00:01:05,010 --> 00:01:09,990 negotiating your next long-term contract with AWS, or just figuring 21 00:01:09,990 --> 00:01:12,060 out why it increasingly resembles of. 22 00:01:12,155 --> 00:01:15,785 Phone number, but nobody seems to quite know why that is. 23 00:01:16,085 --> 00:01:19,655 To learn more, visit duck bill hq.com. 24 00:01:19,955 --> 00:01:22,835 Remember, you can't duck the duck bill. 25 00:01:22,865 --> 00:01:28,265 Bill, which my CEO reliably informs me is absolutely not our slogan. 26 00:01:28,895 --> 00:01:33,545 So I can do a piss week job of explaining and mispronouncing almost anything. 27 00:01:33,545 --> 00:01:36,760 But in the interest of time, why don't you describe what it is you're building. 28 00:01:37,605 --> 00:01:38,054 Hunter Leath: Sure. 29 00:01:38,054 --> 00:01:43,575 So our company Archil, is transforming S3 buckets into Infinite Post six 30 00:01:43,575 --> 00:01:48,134 compatible file systems that provide instant access to massive data sets. 31 00:01:48,255 --> 00:01:50,565 So you can run anything directly on S3. 32 00:01:51,315 --> 00:01:51,975 Corey Quinn: Well jokes on you. 33 00:01:51,975 --> 00:01:55,425 If you improperly mount things and start using it as a file system, then 34 00:01:55,664 --> 00:01:58,845 people have been doing that since S3 launched and then they led to tears 35 00:01:58,845 --> 00:02:03,285 before bedtime when request charges were levied, uh, starting in late beta around 36 00:02:03,285 --> 00:02:07,274 a lot of that just because it caused a lot of problem for the front end stuff. 37 00:02:07,455 --> 00:02:11,954 Uh, AWS has also come out with their Mount point open source nonsense. 38 00:02:12,105 --> 00:02:14,805 So, you know, after 15 years of, sorry, 20 years of telling people, 39 00:02:14,805 --> 00:02:16,785 oh, S3 is not a file system. 40 00:02:16,935 --> 00:02:18,195 Don't use it that way. 41 00:02:18,255 --> 00:02:22,204 They trot out and here's how to use it as a. File system because someone is trying 42 00:02:22,204 --> 00:02:26,644 to see if some customer somewhere is gonna snap and just go after them at reinvent. 43 00:02:26,885 --> 00:02:27,185 Hunter Leath: Yeah. 44 00:02:27,185 --> 00:02:28,685 I mean, I don't know. 45 00:02:28,685 --> 00:02:32,975 You, you may have seen a year ago we posted on the AWS subreddit, uh, talking 46 00:02:32,975 --> 00:02:37,325 about the solution and almost all of the top comments were S3 is not a file system. 47 00:02:37,325 --> 00:02:38,555 This is a terrible idea. 48 00:02:38,555 --> 00:02:42,545 I would never recommend this to my customers, which I love. 49 00:02:42,545 --> 00:02:47,015 I love that kind of feedback because I think what we've done is we found a really 50 00:02:47,015 --> 00:02:49,055 unique approach to solve this problem. 51 00:02:49,575 --> 00:02:53,805 That hasn't been done before by any of these fuse solutions in the past. 52 00:02:54,300 --> 00:02:56,925 Corey Quinn: I, I, I wanna point out one other thing because this, this does 53 00:02:56,925 --> 00:03:00,855 sound to the uninitiated, given things I've said before, like I've somehow 54 00:03:00,855 --> 00:03:05,235 pivoted into becoming a credulous moron because a lot of people have 55 00:03:05,235 --> 00:03:07,035 tried to take a bite at this apple. 56 00:03:07,125 --> 00:03:10,545 But one thing that makes you a little bit different is you spent eight years at 57 00:03:10,545 --> 00:03:12,975 Amazon working on the elastic file system. 58 00:03:13,170 --> 00:03:18,480 EFS product, which generally you can say a lot about Amazon, they don't tend 59 00:03:18,480 --> 00:03:22,619 to hire stupid and they certainly don't suffer it for the better part of a decade. 60 00:03:22,859 --> 00:03:24,630 You clearly understand the space. 61 00:03:24,630 --> 00:03:28,649 You've worked in it, you know where the bodies are more or less buried here, 62 00:03:28,890 --> 00:03:30,899 and you think you see something here. 63 00:03:31,399 --> 00:03:35,119 That alone gets my interest and my attention more so than anyone 64 00:03:35,119 --> 00:03:37,549 starting it like, well, we're gonna build this thing in a product space 65 00:03:37,549 --> 00:03:38,780 we've never actually worked in. 66 00:03:38,780 --> 00:03:40,399 How hard could it really be? 67 00:03:40,700 --> 00:03:44,119 Uh, and for some reason that pitch, if it uses the word AI and it somewhere 68 00:03:44,119 --> 00:03:45,890 raises a $30 million seed round. 69 00:03:45,950 --> 00:03:47,269 Hunter Leath: Yeah, I, I mean, I think that's. 70 00:03:47,785 --> 00:03:52,975 Right, and I really enjoyed my time at AWS on EFS, and it taught 71 00:03:52,975 --> 00:03:56,605 me all kinds of things about running at scale storage solutions. 72 00:03:56,935 --> 00:04:00,595 I think I went home every day for eight years thinking about 73 00:04:00,595 --> 00:04:05,215 the problems associated with EFS and performance and cost and how 74 00:04:05,215 --> 00:04:07,525 to optimize storage placement. 75 00:04:07,585 --> 00:04:14,605 And eventually came to this realization that what EFS does as this. 76 00:04:15,000 --> 00:04:18,630 Scale out file system that you can dump things in is probably the 77 00:04:18,630 --> 00:04:20,490 right shape for a lot of customers. 78 00:04:20,820 --> 00:04:24,810 If we could add that synchronization to S3 and the performance that they 79 00:04:24,810 --> 00:04:28,860 expect, which we think that we can bring to the table, and so rather than being 80 00:04:28,860 --> 00:04:33,780 just a fused library, like you might see S3 Fs or goofy Fs from the past, 81 00:04:33,960 --> 00:04:35,490 Corey Quinn: you're going through user space in those cases and the 82 00:04:35,490 --> 00:04:38,039 latency hit is wildly unacceptable for an awful lot of things. 83 00:04:38,470 --> 00:04:39,040 Hunter Leath: That's right. 84 00:04:39,100 --> 00:04:45,220 We are running this middle layer of NVME devices, which allows us to provide 85 00:04:45,490 --> 00:04:50,860 SSD, like EDS, like access to the data that's already there for our customers. 86 00:04:50,860 --> 00:04:52,675 So they get a file system like experience. 87 00:04:53,685 --> 00:04:54,435 Corey Quinn: Uh, this can be done. 88 00:04:54,435 --> 00:04:58,695 I wanna be clear that back in 2008 at Reach Local, we ran the 89 00:04:58,695 --> 00:05:03,195 entire site on a MySQL database that lived on a NetApp volume. 90 00:05:03,555 --> 00:05:06,045 This can be done over NFS. 91 00:05:06,045 --> 00:05:10,245 There's a lot of dialing in and tweaking and care you have to take. 92 00:05:10,455 --> 00:05:13,815 But if we were able to do it back then, we, this isn't knowledge of the 93 00:05:13,815 --> 00:05:15,435 ancients that has somehow been lost. 94 00:05:15,435 --> 00:05:19,965 You can run these highly critical transactional workloads on something over 95 00:05:19,965 --> 00:05:22,035 a wire that is, that is no longer in. 96 00:05:22,040 --> 00:05:22,580 Doubt. 97 00:05:22,850 --> 00:05:25,490 I found that when EFS launched, it had some challenges. 98 00:05:25,490 --> 00:05:28,610 I was very vocal about them, talked to several people on the team. 99 00:05:28,760 --> 00:05:31,880 As I recall, they were very large at reinvent, had no neck, and I 100 00:05:31,880 --> 00:05:33,560 thought, well, this is the end of me. 101 00:05:33,740 --> 00:05:36,530 But the question was, tell us more about the use case. 102 00:05:36,530 --> 00:05:39,530 And over time, it's become one of my favorite services. 103 00:05:39,710 --> 00:05:43,910 I I, for a while, I was running my home directory on my EC2 box out of EFS. 104 00:05:44,060 --> 00:05:46,910 I had to stop doing that because it took forever to instantiate 105 00:05:46,910 --> 00:05:48,140 a new uh, shell system. 106 00:05:48,605 --> 00:05:53,135 Turns out that every time it read my ZSH profile, it was making 1500 107 00:05:53,135 --> 00:05:55,715 discreet read operations against disc. 108 00:05:55,715 --> 00:05:59,135 And maybe that's not the best use case over the wire, but that's 109 00:05:59,135 --> 00:06:02,705 also 20 years of legacy Croft in my case that I had to worry about. 110 00:06:03,155 --> 00:06:08,375 So this is in AWS, this is living on prem, this is in other cloud environments. 111 00:06:08,705 --> 00:06:10,410 Where where is this thing based? 112 00:06:10,790 --> 00:06:12,710 You said S3, so I have some guesses. 113 00:06:12,800 --> 00:06:17,540 Hunter Leath: We today have shared clusters that we sell to customers, 114 00:06:17,570 --> 00:06:22,610 similar to how S3 or EFS or any of these services work in a couple of regions in 115 00:06:22,610 --> 00:06:27,440 AWS and a single region In GCP, we're also working with customers for some 116 00:06:27,440 --> 00:06:31,850 on-premise deployments, but we want to be where the compute is that uses the 117 00:06:31,850 --> 00:06:34,010 data to reduce that time on the wire. 118 00:06:34,545 --> 00:06:34,755 Corey Quinn: Right. 119 00:06:34,755 --> 00:06:38,775 The latency issue for anything that is being used by, as you said, presents as 120 00:06:38,775 --> 00:06:42,435 an NVME device, that means you've gotta be hooking the kernel somehow, uh, locally on 121 00:06:42,435 --> 00:06:44,505 the system, and there has to be a cache. 122 00:06:44,505 --> 00:06:47,835 You can't be going to S3 for every read and every rite. 123 00:06:47,840 --> 00:06:50,325 The, the, the request charges alone would make this a 124 00:06:50,325 --> 00:06:51,915 complete financial non-starter. 125 00:06:52,275 --> 00:06:53,745 What's the overall architecture look like? 126 00:06:53,860 --> 00:06:58,330 Hunter Leath: Yeah, so what we do is we are running a fleet of servers that 127 00:06:58,330 --> 00:07:00,610 have the SSD instances attached to them. 128 00:07:00,850 --> 00:07:04,720 Similar to, there's like a lot in the news about planet scale, planet 129 00:07:04,720 --> 00:07:09,520 scale metal, very similar thing that we're doing here, and those 130 00:07:09,520 --> 00:07:11,620 servers provide this distributed. 131 00:07:12,150 --> 00:07:16,350 Durable cache for data that's going into or out of S3. 132 00:07:16,740 --> 00:07:21,030 And what that allows us to do is provide a very easy on-ramp for people 133 00:07:21,030 --> 00:07:25,710 who want to use S3 as a file system or use it as their backing store, but 134 00:07:25,710 --> 00:07:30,150 then also give customers this ability to tell us, well, maybe we actually 135 00:07:30,480 --> 00:07:34,770 only want this data to be stored on SSDs for performance or cost reasons. 136 00:07:34,830 --> 00:07:37,170 And we give that flexibility to our users. 137 00:07:37,740 --> 00:07:40,950 Corey Quinn: Now the right way a lot of these workloads have been done is you 138 00:07:40,950 --> 00:07:46,080 teach your workloads how to use object store and teach at those semantics and it 139 00:07:46,080 --> 00:07:48,390 grabs whatever it needs and life goes on. 140 00:07:48,570 --> 00:07:53,099 That is a wonderful end state, but there's a universe of applications out there 141 00:07:53,099 --> 00:07:54,510 that flat out are not built that way. 142 00:07:54,840 --> 00:07:58,830 Where, oh, just teach your 30-year-old application how object store works. 143 00:07:58,860 --> 00:07:59,610 Oh, terrific. 144 00:07:59,610 --> 00:08:02,310 After you please, if it's not that hard, I like. 145 00:08:02,325 --> 00:08:03,315 Due to meet my mainframe. 146 00:08:03,735 --> 00:08:05,685 There are a bunch of legacy workloads. 147 00:08:05,685 --> 00:08:07,604 I can see where this is incredibly helpful. 148 00:08:07,875 --> 00:08:10,305 Do you see modern workloads taking advantage as well? 149 00:08:10,395 --> 00:08:11,295 Hunter Leath: I agree. 150 00:08:11,295 --> 00:08:14,655 I think it's a combination and, and I've had conversations with people in 151 00:08:14,655 --> 00:08:19,515 the past where I kind of like correct the legacy term because in many of 152 00:08:19,515 --> 00:08:23,594 these fields it's, it's not even that these applications are necessarily old. 153 00:08:23,594 --> 00:08:28,305 Like if you look at the HPC and academic fields, these are new things 154 00:08:28,305 --> 00:08:31,695 that are being written that are the best in class for what they do. 155 00:08:32,655 --> 00:08:36,584 But are built against file systems because most universities and labs 156 00:08:36,584 --> 00:08:41,265 are running on some kind of shared cluster mainframe thing with a 157 00:08:41,294 --> 00:08:42,885 file system sitting on the side. 158 00:08:43,275 --> 00:08:49,395 And so I agree that we provide an opportunity to take this vast library of 159 00:08:49,395 --> 00:08:54,135 software that exists today and actually connect it to the cloud in a cloud native 160 00:08:54,135 --> 00:08:56,234 way so that the data could be used by. 161 00:08:57,660 --> 00:08:59,370 Lamb from someone on a different team. 162 00:08:59,520 --> 00:09:04,770 But I also think that we are going to see a resurgence in 163 00:09:05,010 --> 00:09:06,295 applications that are built. 164 00:09:07,110 --> 00:09:12,750 In a modern way, four file systems, and my belief is that people 165 00:09:12,750 --> 00:09:18,720 have shifted development to using object storage natively because 166 00:09:18,720 --> 00:09:20,670 nothing like Archil has existed. 167 00:09:21,060 --> 00:09:25,860 And traditionally, the limits of the file system have been so painful for people. 168 00:09:26,100 --> 00:09:30,390 If you've run a NFS share and you know, run out of space or run out 169 00:09:30,390 --> 00:09:34,110 of iops or gotten paged at 4:00 AM it leaves this bad taste in your 170 00:09:34,110 --> 00:09:35,940 mouth about the entire technology. 171 00:09:36,315 --> 00:09:41,385 Where it turns out, we can build file systems that scale in a very 172 00:09:41,385 --> 00:09:45,165 similar way to S3 and provide a very similar experience. 173 00:09:45,675 --> 00:09:49,605 Corey Quinn: The S3 bucket that's backing this has to be, have the data laid down 174 00:09:49,605 --> 00:09:54,015 in it in a particular format that you folks, uh, uh, work with, or it can be 175 00:09:54,015 --> 00:09:55,575 hooked up to existing large data sets. 176 00:09:55,695 --> 00:09:58,545 Hunter Leath: It's the latter, and that's what we are super excited about, 177 00:09:58,545 --> 00:10:00,705 is that if you are a user who has. 178 00:10:00,835 --> 00:10:05,965 An immense quantity of geospatial data or video data, or any of these data 179 00:10:05,965 --> 00:10:07,615 sets that you might need to process. 180 00:10:07,765 --> 00:10:11,575 We work with that data directly and we synchronize in both directions, 181 00:10:11,575 --> 00:10:15,325 so anything that you put into our file system then gets replicated 182 00:10:15,325 --> 00:10:17,120 into S3 and its original format too. 183 00:10:17,954 --> 00:10:19,305 And you get to own that data. 184 00:10:19,680 --> 00:10:22,365 Corey Quinn: You, you talk about, uh, shareable across multiple 185 00:10:22,365 --> 00:10:24,105 instances simultaneously for datasets. 186 00:10:24,135 --> 00:10:26,834 Um, I, I have to imagine that there are some race condition 187 00:10:26,865 --> 00:10:28,605 and concerning limits here. 188 00:10:28,665 --> 00:10:32,535 You, to use my, my insane example, you can't necessarily, I would 189 00:10:32,535 --> 00:10:36,555 think, have a MySQL database living in S3 that you then wind up pulling 190 00:10:36,555 --> 00:10:40,214 out, having multiple things, commit transactions to that thing, and then 191 00:10:40,214 --> 00:10:41,685 put it back and hope for the best. 192 00:10:41,685 --> 00:10:43,964 I, but how does the, how does the fencing work? 193 00:10:44,230 --> 00:10:47,230 Hunter Leath: That's also a very good question and something that stumped us 194 00:10:47,230 --> 00:10:51,790 early on around how we actually extend mutability to multiple instances. 195 00:10:52,210 --> 00:10:56,200 But we, we started from this premise that many customers who 196 00:10:56,200 --> 00:10:57,395 are running workloads in the cloud. 197 00:10:58,109 --> 00:11:03,479 We'll pick something like EBS as an easy starting point to store their data and 198 00:11:03,479 --> 00:11:07,560 then only move to shared storage when they find out that for high availability 199 00:11:07,560 --> 00:11:12,900 reasons, all the data can't be attached to a single instance, for example, and 200 00:11:12,930 --> 00:11:18,180 we want to build almost a version of EBS that doesn't come with any performance 201 00:11:18,180 --> 00:11:23,339 drawbacks, but allows you to have this sharing across instances in a safe way. 202 00:11:23,910 --> 00:11:29,970 Has the synchronization to S3 and does the auto-scaling and the pay per usage that 203 00:11:29,970 --> 00:11:33,360 people like from services like S3 and EFS. 204 00:11:33,569 --> 00:11:39,329 But ultimately the application ends up being the one responsible for 205 00:11:39,329 --> 00:11:41,310 coordinating rights to that data. 206 00:11:41,579 --> 00:11:46,800 So we would not recommend that if you launch a MySQL database, you then 207 00:11:46,980 --> 00:11:50,579 attach multiple instances to that database and write to it simultaneously. 208 00:11:50,925 --> 00:11:53,025 Because MySQL isn't designed for that. 209 00:11:53,535 --> 00:11:54,645 Corey Quinn: Well, not with that attitude. 210 00:11:55,695 --> 00:11:55,935 Sorry. 211 00:11:55,935 --> 00:11:56,655 Please continue. 212 00:11:56,655 --> 00:12:00,585 I, I misuse databases because I have problems that I make everyone else's. 213 00:12:00,765 --> 00:12:02,055 Hunter Leath: You can do it once, right? 214 00:12:02,055 --> 00:12:03,195 Like many things, you can do it. 215 00:12:03,195 --> 00:12:03,675 One time. 216 00:12:04,005 --> 00:12:06,685 Corey Quinn: It worked in, it worked in my laptop, so off to production with it. 217 00:12:06,685 --> 00:12:06,790 Why not? 218 00:12:07,305 --> 00:12:09,195 That's why I call my staging environment theory works. 219 00:12:09,195 --> 00:12:10,245 In theory, not in production, 220 00:12:10,425 --> 00:12:13,305 Hunter Leath: but for many workloads that are maybe producing files or 221 00:12:13,305 --> 00:12:17,145 like transcoding videos and have a multi-stage pipeline where different 222 00:12:17,145 --> 00:12:19,395 instances are handing off that video file. 223 00:12:19,665 --> 00:12:22,875 Being able to share that data across instances is very helpful. 224 00:12:23,445 --> 00:12:26,415 Corey Quinn: Oh, I can absolutely see where that, that makes sense. 225 00:12:26,445 --> 00:12:30,885 The, the pricing, at least as on your website, is also compelling. 226 00:12:31,155 --> 00:12:33,615 Uh, does the data live in the user's S3 bucket? 227 00:12:33,615 --> 00:12:34,785 Does it live in yours? 228 00:12:34,875 --> 00:12:36,495 Where is the, where does that. 229 00:12:37,020 --> 00:12:38,640 Live, for lack of a better term. 230 00:12:39,030 --> 00:12:43,920 Hunter Leath: Yeah, so we attach to customers S3 buckets directly, so we use 231 00:12:43,920 --> 00:12:45,360 the data that's already in their bucket. 232 00:12:45,390 --> 00:12:49,349 We synchronize any changes back to their bucket and they get to keep that data. 233 00:12:49,560 --> 00:12:54,180 We're working to simplify onboarding so customers can try us without maybe 234 00:12:54,180 --> 00:12:58,170 making an S3 bucket or attaching us to their Productionist three bucket. 235 00:12:58,349 --> 00:13:01,770 And so we're actually launching the ability to use a bucket that we manage. 236 00:13:02,175 --> 00:13:04,275 Uh, so that you don't have to worry about any of that if 237 00:13:04,275 --> 00:13:05,565 you're not coming in with data. 238 00:13:06,045 --> 00:13:08,175 Corey Quinn: Something that strikes me about this is your pricing. 239 00:13:08,385 --> 00:13:13,540 The first thing I always look for on a website is the, it's 240 00:13:13,540 --> 00:13:16,365 the pricing page, because that tells me who is this actually for? 241 00:13:16,485 --> 00:13:18,765 And you generally want two things. 242 00:13:18,765 --> 00:13:21,285 One is the get started right now because it's two in the 243 00:13:21,285 --> 00:13:22,275 morning and I have a problem. 244 00:13:22,275 --> 00:13:25,185 And if it's call here to contact us, you're not going to space today. 245 00:13:25,245 --> 00:13:29,115 And the other is the enterprise end of it, which is contact us. 246 00:13:29,115 --> 00:13:29,305 We don't. 247 00:13:29,615 --> 00:13:33,965 Put our data with random websites because we are a Fortune 500 and strive 248 00:13:33,965 --> 00:13:38,015 to act like one and whatever's in between them almost doesn't matter. 249 00:13:38,225 --> 00:13:39,815 And you have two options right now. 250 00:13:39,815 --> 00:13:44,615 One is the enterprise option and the other is the developer plan at 20 cents a month. 251 00:13:44,675 --> 00:13:50,345 And that is a very resonant figure for me because 10 cents a month gets 252 00:13:50,345 --> 00:13:53,405 you single availability zone EFS at. 253 00:13:53,584 --> 00:14:00,064 On demand prices and 30 cents a month gets you multi a z uh, EFS, you're smack 254 00:14:00,064 --> 00:14:04,324 dab in the middle of them with a better durability story than EFS itself offers. 255 00:14:04,714 --> 00:14:08,735 And even with having to store the original data on, in S3 standard as 256 00:14:08,735 --> 00:14:12,785 well, which is another two and a half cents, you are, we're still talking 257 00:14:12,814 --> 00:14:17,405 that this is less expensive for active data than using EFS off the shelf. 258 00:14:17,435 --> 00:14:17,704 Hunter Leath: Yeah. 259 00:14:17,704 --> 00:14:20,345 And that, that was the goal of ours was to. 260 00:14:20,445 --> 00:14:25,365 Be in a position to offer this experience like EFS without having to charge the 261 00:14:25,365 --> 00:14:29,985 prices that I know so many customers are frustrated with from a thing like EFS. 262 00:14:30,675 --> 00:14:33,195 And I think too, if you, if you zoom out further. 263 00:14:33,540 --> 00:14:37,380 And you talk to users of AWS who are not familiar with EFS. 264 00:14:37,380 --> 00:14:40,320 What they compare it to is EBS pricing. 265 00:14:40,830 --> 00:14:45,300 And when they see 20 cents a gigabyte, it looks extremely large compared 266 00:14:45,300 --> 00:14:49,020 to the 8 cents a gigabyte that you get for provisioned GP3 volume. 267 00:14:50,090 --> 00:14:54,319 But we've actually run the math, and if you do an apples to apples 268 00:14:54,319 --> 00:14:58,850 comparison of taking data that was previously on EBS and moving it to our 269 00:14:58,850 --> 00:15:03,590 service, the pricing actually clocks in at something like 1.95 cents. 270 00:15:03,930 --> 00:15:07,560 Per provision gigabyte that you would have had on EBS. 271 00:15:08,040 --> 00:15:08,250 Corey Quinn: Right? 272 00:15:08,250 --> 00:15:11,580 Because you have to over, you have to over-provision an EBS volume because 273 00:15:11,580 --> 00:15:14,670 only a lunatic runs at a hundred percent, uh, utilization on it. 274 00:15:15,000 --> 00:15:17,640 You have to monitor that and care about it. 275 00:15:17,640 --> 00:15:21,270 And people from the EFS side will say, well, if you're not using the data, 276 00:15:21,270 --> 00:15:24,480 we have intelligent tiering, which makes it a lot less per gigabyte. 277 00:15:24,630 --> 00:15:24,870 Yeah. 278 00:15:24,870 --> 00:15:27,720 You only charge per active gigabyte in the course of a month there 279 00:15:27,720 --> 00:15:29,070 at, I believe, a 30 day cycle. 280 00:15:29,220 --> 00:15:31,005 So yeah, you, you drop to zero. 281 00:15:31,375 --> 00:15:36,025 Where they drop down to, I forget the exact number, but it's, it is compelling. 282 00:15:36,265 --> 00:15:39,445 The economics alone are fascinating. 283 00:15:40,165 --> 00:15:42,355 All you have to worry about after that is okay. 284 00:15:42,355 --> 00:15:45,175 Does, does the thing actually work for a given use case? 285 00:15:45,385 --> 00:15:46,225 Hunter Leath: Yeah, that's right. 286 00:15:46,225 --> 00:15:51,025 And it is rare, I think, to find these kinds of optimizations and infrastructure 287 00:15:51,025 --> 00:15:53,065 where you can both save people money. 288 00:15:53,400 --> 00:15:55,290 And make their software faster. 289 00:15:55,560 --> 00:15:59,250 And we think that this is one of those examples where that's possible, where 290 00:15:59,250 --> 00:16:04,140 if you're moving from something like an All SSD, an EBS or an EFS deployment, 291 00:16:04,350 --> 00:16:07,980 we can save you significant money on top of that existing deployment. 292 00:16:08,400 --> 00:16:12,060 And if you're moving from something like using S3 directly or downloading 293 00:16:12,060 --> 00:16:16,050 a zip file and then zipping it, we can make that workload faster because we're 294 00:16:16,050 --> 00:16:18,420 actually keeping some of the data on SSDs. 295 00:16:18,810 --> 00:16:20,850 But yes, I think we. 296 00:16:21,855 --> 00:16:25,785 Plan to do a better job over the coming months of publishing more benchmarks 297 00:16:25,785 --> 00:16:30,885 and, and more customers who are using the service to highlight the applications 298 00:16:30,885 --> 00:16:34,995 that work well with Archil so that people know that it's an easy thing to adopt 299 00:16:35,475 --> 00:16:37,785 Corey Quinn: if the pull through cash, more or less, with really 300 00:16:37,785 --> 00:16:38,955 great performance characteristics. 301 00:16:39,345 --> 00:16:43,665 So a concern I have historically, I did a piece back in 20 19, 20 18, 302 00:16:44,025 --> 00:16:46,605 about how to compete with Amazon and. 303 00:16:47,160 --> 00:16:50,655 One of 'em obviously is good developer experience because that is something 304 00:16:50,655 --> 00:16:54,824 that has not gone super well for them, uh, in recent years compared to 305 00:16:54,824 --> 00:16:58,454 a number of other folks, but also to find things that aren't on the same 306 00:16:58,454 --> 00:16:59,895 rails that they're innovating on. 307 00:17:00,074 --> 00:17:06,585 This feels like it could be a story of a feature that AWS puts out at some point, 308 00:17:07,004 --> 00:17:09,780 and at that if they wind up effectively. 309 00:17:10,085 --> 00:17:13,595 Fronting S3 with an EFS, like cash or something like that. 310 00:17:13,895 --> 00:17:17,525 How does that change your competitive positioning? 311 00:17:17,525 --> 00:17:19,785 I mean, I have to imagine this is come up as an idea. 312 00:17:20,730 --> 00:17:24,270 This episode is sponsored by my own company, duck Bill. 313 00:17:24,540 --> 00:17:28,080 Having trouble with your AWS bill, perhaps it's time to 314 00:17:28,080 --> 00:17:30,180 renegotiate a contract with them. 315 00:17:30,510 --> 00:17:35,910 Maybe you're just wondering how to predict what's going on in the wide world of AWS. 316 00:17:35,970 --> 00:17:38,580 Well, that's where Duck Bill comes to help. 317 00:17:38,790 --> 00:17:41,550 Remember, you can't duck the duck bill. 318 00:17:41,550 --> 00:17:45,150 Bill, which I am reliably informed by my business partner 319 00:17:45,270 --> 00:17:47,760 is absolutely not our motto. 320 00:17:47,815 --> 00:17:51,070 To learn more, visit doc bill hq.com. 321 00:17:51,929 --> 00:17:53,790 Hunter Leath: Honestly, Corey, I, I hope they do. 322 00:17:53,790 --> 00:17:57,659 I have many friends still who are on the EFS service, and I would be ecstatic if 323 00:17:57,659 --> 00:18:02,939 they were able to build something kind of as exciting as synchronizing to S3 because 324 00:18:02,939 --> 00:18:04,590 I think it's important for customers. 325 00:18:04,860 --> 00:18:11,100 But I think that what we are building goes beyond that, which is that AWS, 326 00:18:11,189 --> 00:18:16,889 at least from my experience, is very focused on how to build building 327 00:18:16,889 --> 00:18:18,570 blocks and give them to customers. 328 00:18:18,570 --> 00:18:19,034 And I think even. 329 00:18:20,160 --> 00:18:24,270 Peter DeSantis said, uh, reinvent, uh, several years ago, said something to the 330 00:18:24,270 --> 00:18:26,130 effect of we will not build frameworks. 331 00:18:26,190 --> 00:18:28,050 We want customers to build frameworks. 332 00:18:28,590 --> 00:18:36,060 And so our ability to help our customers is going to be based around how we tie 333 00:18:36,090 --> 00:18:41,070 things tightly together so that they don't have to cobble a bunch of AWS services 334 00:18:41,190 --> 00:18:42,660 in order to get the same solution. 335 00:18:43,110 --> 00:18:46,680 And the way that we do that is through a combination of performance work. 336 00:18:47,324 --> 00:18:50,084 Our goal is actually not to be as fast as EFS. 337 00:18:50,084 --> 00:18:54,254 It's to be as fast as EBS so that customers can replace EBS volumes 338 00:18:54,254 --> 00:18:58,064 with us, which we think is an entirely new market segment that 339 00:18:58,064 --> 00:19:01,245 Amazon won't be able to capture. 340 00:19:01,754 --> 00:19:04,155 Corey Quinn: When you say Eeb, can I compete with EBS? 341 00:19:04,155 --> 00:19:04,514 Does that. 342 00:19:04,525 --> 00:19:08,515 Compete with the bring money IO2 volumes as well, performance wise. 343 00:19:09,175 --> 00:19:11,665 'cause every time someone talks to me about those, it's like, oh great. 344 00:19:11,695 --> 00:19:14,545 Uh, and everything they say after that turns into, and here's how 345 00:19:14,545 --> 00:19:17,305 you build your own sand from Popsicle sticks in the cloud. 346 00:19:17,305 --> 00:19:18,445 Which great. 347 00:19:18,475 --> 00:19:19,885 Not really what I wanna do. 348 00:19:20,035 --> 00:19:20,245 Hunter Leath: Yeah. 349 00:19:20,245 --> 00:19:21,985 I think in time we will. 350 00:19:22,300 --> 00:19:29,230 Uh, for now we see us as like a very good alternative to GP3, which are where 351 00:19:29,230 --> 00:19:33,399 we expect the majority of workloads, which are not transactional database 352 00:19:33,399 --> 00:19:36,940 workloads running, where if you're doing video transcoding, you're probably 353 00:19:36,940 --> 00:19:41,590 using a GP3 volume to store that data and not like a finely tuned IO2 volume. 354 00:19:42,255 --> 00:19:42,525 Corey Quinn: Oh yeah. 355 00:19:42,525 --> 00:19:46,515 Every time I see an IO2 volume, I have many questions, uh, most of which are the 356 00:19:46,515 --> 00:19:50,115 answers become, uh, maybe you shouldn't be using this volume for this use case. 357 00:19:50,325 --> 00:19:57,885 I also, normally, I tend to be relatively down on the idea of AWS 358 00:19:57,885 --> 00:20:00,825 trying to compete with someone else who's making money somewhere because 359 00:20:00,825 --> 00:20:02,270 their developer experience is crappy. 360 00:20:02,370 --> 00:20:05,865 But the, the lockin halo effect of the ecosystem does become challenging. 361 00:20:05,925 --> 00:20:08,895 Um, especially if we take a look at something on the storage 362 00:20:08,895 --> 00:20:11,055 side as an easy example where. 363 00:20:11,475 --> 00:20:15,645 Staying first party, even if it's crappier from an experience, is a lot 364 00:20:15,645 --> 00:20:19,515 more justifiable for folks than bringing something into critical path with 365 00:20:19,515 --> 00:20:21,345 something as fundamental as disc access. 366 00:20:21,435 --> 00:20:25,035 Hunter Leath: You had mentioned to me at some point previously that uh, the 367 00:20:25,035 --> 00:20:29,145 file system is not the place where you want to be betting on new technology, 368 00:20:29,355 --> 00:20:31,215 and I think that's absolutely true. 369 00:20:31,815 --> 00:20:34,335 You know, these, these things are not developed frequently 370 00:20:34,575 --> 00:20:36,015 because there is so much. 371 00:20:36,255 --> 00:20:41,295 Thought and care and safety that needs to go into these critical components. 372 00:20:41,475 --> 00:20:48,795 And so I think that as a business, we are going to have to build our way up to some 373 00:20:48,795 --> 00:20:53,235 of the enterprises that would otherwise be happy to pick an AWS solution. 374 00:20:53,565 --> 00:20:57,545 But if you look at the vast, at just the L. Of storage 375 00:20:57,545 --> 00:20:58,745 workloads that are out there. 376 00:20:59,075 --> 00:21:03,305 There are lots of startups who are interested or need to capture these 377 00:21:03,305 --> 00:21:06,965 marginal cost savings or marginal performance savings in order to win in 378 00:21:06,965 --> 00:21:09,395 their product category where we can start. 379 00:21:09,635 --> 00:21:13,055 And there are also lots of enterprises who have enormous 380 00:21:13,055 --> 00:21:15,035 shared caches for things like. 381 00:21:15,385 --> 00:21:20,755 Docker images or CICD artifacts where if things did go wrong, 382 00:21:20,815 --> 00:21:22,345 it's not a huge loss for them. 383 00:21:22,825 --> 00:21:28,705 And so being able to build our customer base from these easier to win people who 384 00:21:28,705 --> 00:21:33,655 are more open to tech, to new technology and this place workloads and work our way 385 00:21:33,655 --> 00:21:38,815 up to an established member of the storage community like A ZFS or an XFS, is the 386 00:21:38,815 --> 00:21:39,875 path that I think we're going to take. 387 00:21:41,160 --> 00:21:43,650 Corey Quinn: One question I have because you have, uh, you announced 388 00:21:43,650 --> 00:21:47,940 at the top of your website that you've raised a $6.7 million seed round. 389 00:21:47,940 --> 00:21:48,780 Congratulations. 390 00:21:49,080 --> 00:21:53,220 Uh, but what I don't see on this, in, at least above the fold, and as I 391 00:21:53,220 --> 00:21:55,290 scroll down, I don't see it here either. 392 00:21:55,620 --> 00:21:57,510 I see no mention of ai. 393 00:21:57,720 --> 00:21:59,670 So first is that legal? 394 00:22:00,020 --> 00:22:04,730 It toward the end of 2025 to raise money without an AI story front and center 395 00:22:04,730 --> 00:22:06,350 sucking all the oxygen out of the room. 396 00:22:06,620 --> 00:22:08,540 So what is your AI story? 397 00:22:08,720 --> 00:22:10,580 Hunter Leath: Yeah, Corey, I can't believe you've brought this up. 398 00:22:10,580 --> 00:22:14,750 And my investors are gonna find out that we, we've not placed AI above the fold. 399 00:22:15,290 --> 00:22:16,580 I, I think that. 400 00:22:17,040 --> 00:22:22,230 What is exciting about our technology is that it is so horizontal and 401 00:22:22,230 --> 00:22:27,030 broad that it's applicable to all of these workloads that exist today. 402 00:22:27,270 --> 00:22:30,930 Like we talked about video transcoding, CICD, geospatial, these things that 403 00:22:30,930 --> 00:22:33,240 were in the world prior to 2025. 404 00:22:33,810 --> 00:22:38,310 But I also think that there's a lot of interesting things happening in the AI 405 00:22:38,310 --> 00:22:40,800 space that's going to intersect with us. 406 00:22:40,980 --> 00:22:46,170 And what I hear most frequently from customers is that these models are being. 407 00:22:46,335 --> 00:22:52,065 Trained on these data sets that include an immense amount of like 408 00:22:52,065 --> 00:22:54,765 Unix tool usage and file system usage. 409 00:22:55,125 --> 00:22:59,445 And so there are many companies out there that are trying to connect LLMs 410 00:22:59,445 --> 00:23:04,935 to, you know, slack or Salesforce through these purpose built tools. 411 00:23:05,085 --> 00:23:10,005 But the models actually perform better if you're able to expose data in a file 412 00:23:10,005 --> 00:23:11,715 system that the model can just cre. 413 00:23:12,630 --> 00:23:15,900 Corey Quinn: Oh yeah, the, the command line tools are the best MCP out there. 414 00:23:15,900 --> 00:23:18,690 It can do basically anything you need it to do, and that knowledge 415 00:23:18,690 --> 00:23:21,480 of how to use these things is baked into the foundation model itself. 416 00:23:21,690 --> 00:23:25,800 Hunter Leath: And so I think it's my hope that we can become. 417 00:23:26,490 --> 00:23:31,260 As the file system was originally intended to be this universal access layer for 418 00:23:31,260 --> 00:23:37,320 data, no matter where it lives, and then allowing AI models, applications, 419 00:23:37,320 --> 00:23:41,699 agents, what have you, to use our system to access that data is going to allow 420 00:23:41,699 --> 00:23:45,600 them to be more efficient and more productive than they otherwise would be. 421 00:23:46,139 --> 00:23:47,879 Corey Quinn: When you tell the story of what you're building 422 00:23:47,879 --> 00:23:50,580 to folks, what are some of those common misunderstandings you see? 423 00:23:51,225 --> 00:23:55,905 Hunter Leath: Our product is complex as, as you can probably tell from 424 00:23:55,905 --> 00:23:57,055 the conversation that we've had. 425 00:23:57,929 --> 00:24:01,590 Corey Quinn: It's one of those things that feels extraordinarily simple and never is. 426 00:24:01,649 --> 00:24:05,850 Hunter Leath: So we have folks who come in and aren't sure if we are here to 427 00:24:05,850 --> 00:24:09,600 save them money on their S3 bill, which would be great, but of course we are 428 00:24:09,600 --> 00:24:12,840 not, and they're not sure necessarily. 429 00:24:13,445 --> 00:24:15,784 Corey Quinn: Technically, if you're treating it like a file system and just 430 00:24:15,784 --> 00:24:20,165 have her in the crap out of request, yes, yes, we will, but you're doing something 431 00:24:20,165 --> 00:24:24,935 wrong that's right up there with help cut your AWS, uh, by bill by 98% by no longer 432 00:24:24,935 --> 00:24:26,645 checking your credentials into GitHub. 433 00:24:27,665 --> 00:24:29,825 Hunter Leath: Yeah, yeah, exactly. 434 00:24:30,245 --> 00:24:34,595 And so I, I think that we have this, this narrow focus today. 435 00:24:34,935 --> 00:24:39,495 On how we can help people who are either working in the AI space like we talked 436 00:24:39,495 --> 00:24:44,025 about, or have these applications which are built around file system semantics and 437 00:24:44,025 --> 00:24:49,515 connecting those applications to a data lake or downstream object storage native 438 00:24:49,605 --> 00:24:53,475 systems where we will win ultimately. 439 00:24:54,195 --> 00:24:57,885 Corey Quinn: I, I think that that is probably a, it's a user education problem. 440 00:24:58,155 --> 00:25:01,365 There's, which is kind of a, a testament to how reliable storage 441 00:25:01,365 --> 00:25:02,415 has gotten across the board. 442 00:25:02,535 --> 00:25:05,805 This is something that it, it takes a relatively narrow subset of the 443 00:25:05,805 --> 00:25:09,075 industry to actually even think or care about these things, but 444 00:25:09,075 --> 00:25:11,355 those who do really care a lot. 445 00:25:11,680 --> 00:25:16,180 Bout it, it, it's also kind of wild that we've watched EC2 alone get to a point 446 00:25:16,180 --> 00:25:22,990 where we can trust it now for things of this level of latency requirement as well 447 00:25:22,990 --> 00:25:26,050 as durability, the world has changed. 448 00:25:26,140 --> 00:25:26,380 Hunter Leath: Yeah. 449 00:25:26,380 --> 00:25:30,220 I, and there's like a tremendous amount of engineering effort and, 450 00:25:30,490 --> 00:25:35,110 uh, work that goes into making EC2 a durable service for sure. 451 00:25:35,350 --> 00:25:38,080 But it is funny that you say that about people caring. 452 00:25:38,140 --> 00:25:42,730 Uh, because a week or two ago, I think I reached out to the CTO of a company 453 00:25:43,000 --> 00:25:45,880 doing 200, $300 million in annual revenue. 454 00:25:46,679 --> 00:25:49,830 And he picked up the phone immediately and told me that this is a problem that 455 00:25:49,830 --> 00:25:51,929 he's been thinking about for six years. 456 00:25:52,200 --> 00:25:56,490 Because for, for some segment of storage workloads, they're just woefully 457 00:25:56,580 --> 00:25:58,590 underserved by the offerings today. 458 00:25:58,590 --> 00:26:05,040 Whether it's an Amazon offering or a pure WCA vast, uh, NetApp, none of these things 459 00:26:05,040 --> 00:26:10,560 have the right latency and throughput characteristics for all workloads, and 460 00:26:10,560 --> 00:26:14,159 that's why we see such a vast array of storage options that you can buy. 461 00:26:14,925 --> 00:26:18,284 Corey Quinn: It's, I, I can't find it at the moment, but there, the EFS site 462 00:26:18,284 --> 00:26:22,365 used to say with their intelligent tiering offering, uh, what percentage 463 00:26:22,365 --> 00:26:26,294 of data was, uh, there but not accessed in the course of a calendar month. 464 00:26:26,415 --> 00:26:31,725 And it was something north of 80% in the common case, which is effectively some of 465 00:26:31,725 --> 00:26:33,495 the best marketing you could do for this. 466 00:26:33,554 --> 00:26:34,885 The challenge is, is great, but. 467 00:26:35,215 --> 00:26:36,355 I do need to access some of it. 468 00:26:36,355 --> 00:26:39,175 I never really know which one, and it has to be capable of 469 00:26:39,175 --> 00:26:40,615 being reached via a file system. 470 00:26:40,615 --> 00:26:41,245 Semantic. 471 00:26:41,485 --> 00:26:44,905 Well, this sounds like a relatively great way to get started. 472 00:26:45,205 --> 00:26:48,625 It does not sound from the try here to get started. 473 00:26:48,655 --> 00:26:51,205 Uh, documentation you have that it's a particularly heavy lift to 474 00:26:51,205 --> 00:26:52,855 integrate into a test environment. 475 00:26:53,670 --> 00:26:56,760 This is, it's, I wouldn't, I didn't expect to see a lot of 476 00:26:56,760 --> 00:27:03,330 innovation in the block storage world in the year of our Lord 2025. 477 00:27:03,390 --> 00:27:04,470 And yet here we are. 478 00:27:04,620 --> 00:27:04,890 Hunter Leath: Yeah. 479 00:27:04,895 --> 00:27:08,250 And, and that's what I, I go around telling people is funny is that we're 480 00:27:08,250 --> 00:27:12,090 in San Francisco doing a startup in 2025 and we're building a file system. 481 00:27:12,360 --> 00:27:12,630 Corey Quinn: Sorry. 482 00:27:12,630 --> 00:27:15,420 It's a file level semantics and not, uh, block level semantics. 483 00:27:15,420 --> 00:27:17,220 I, I could, it feels like it's right on the cusp. 484 00:27:17,340 --> 00:27:17,880 Hunter Leath: Yes. 485 00:27:18,135 --> 00:27:21,270 I, I think that you are correct. 486 00:27:21,795 --> 00:27:26,835 In that one of these learnings that I had being focused on EFS 487 00:27:26,835 --> 00:27:29,565 and talking to customers is that. 488 00:27:30,345 --> 00:27:35,235 Many people don't view the file system as a place where they want their data 489 00:27:35,235 --> 00:27:39,705 to live long term or, or they don't like the idea of having to pick, if I'm going 490 00:27:39,705 --> 00:27:44,715 to store this data on a NetApp file system forever and lock it away from the 491 00:27:44,715 --> 00:27:48,735 rest of the ecosystem that's evolving around object storage in the cloud. 492 00:27:49,280 --> 00:27:54,080 And really what, what we believe people want is this ability to use 493 00:27:54,080 --> 00:27:56,240 the file system to access that data. 494 00:27:56,480 --> 00:27:59,629 And whenever they're using the data, it makes sense to use it through a 495 00:27:59,629 --> 00:28:03,620 file system, but when they're not using it, they want it to just live in S3. 496 00:28:03,740 --> 00:28:07,070 So it can be like any other piece of data that they're keeping track of. 497 00:28:07,600 --> 00:28:10,629 Corey Quinn: I I do have to ask, what is the response? 498 00:28:10,629 --> 00:28:14,229 If some, if a piece of data lives in S3 via intelligent tiering, for example, 499 00:28:14,320 --> 00:28:18,639 and it's been aged out into some of the, uh, archive tier where it will 500 00:28:18,639 --> 00:28:22,719 take some time, often measured with the calendar in order to come back. 501 00:28:22,780 --> 00:28:24,729 Uh, does it just blow itself the chunks? 502 00:28:24,729 --> 00:28:27,699 Does it just read a file, read error in some way that gets 503 00:28:27,699 --> 00:28:29,260 surfaced to the file system? 504 00:28:29,260 --> 00:28:32,350 That this is not a catastrophic event, but try later? 505 00:28:32,409 --> 00:28:32,649 Hunter Leath: Yeah. 506 00:28:32,649 --> 00:28:35,320 I, today what we do is we just. 507 00:28:35,909 --> 00:28:39,929 Basically wait, which obviously isn't ideal for a lot of applications, 508 00:28:39,959 --> 00:28:43,080 but I think this is one of those give and takes with file systems 509 00:28:43,139 --> 00:28:44,370 that we mentioned earlier. 510 00:28:44,580 --> 00:28:49,590 There's not an easy way for us to signal via Linux that some piece of data is maybe 511 00:28:49,590 --> 00:28:52,260 there but not immediately accessible. 512 00:28:52,590 --> 00:28:57,419 And obviously I think we're going to spend time with our customers to understand as 513 00:28:57,419 --> 00:29:01,830 we hit people who are using these Glacier Deep Archive kind of storage classes, 514 00:29:02,010 --> 00:29:03,959 what kind of experience they would prefer. 515 00:29:04,605 --> 00:29:06,855 Then surface that to them in a usable way. 516 00:29:07,185 --> 00:29:08,835 Corey Quinn: Do you present as a custom file system? 517 00:29:08,835 --> 00:29:11,295 Are you, uh, presenting as EXG something else? 518 00:29:11,295 --> 00:29:12,825 Hunter Leath: We are a custom file system. 519 00:29:12,855 --> 00:29:15,675 We are based on, we run infuse right now, but we're working our 520 00:29:15,675 --> 00:29:17,175 way into the kernel over time. 521 00:29:17,655 --> 00:29:18,045 Corey Quinn: Excellent. 522 00:29:18,075 --> 00:29:20,925 Uh, via module or are you dreams of one day being mainlined? 523 00:29:20,955 --> 00:29:23,895 Hunter Leath: I absolutely, I think that we should be mainlined. 524 00:29:23,925 --> 00:29:28,395 I think that we see development as this iterative step. 525 00:29:28,814 --> 00:29:34,935 Towards stability and what Fuse offers us is an ability to develop very rapidly 526 00:29:34,935 --> 00:29:38,774 and try different things because we're not messing with the kernel that's going 527 00:29:38,774 --> 00:29:42,419 to eventually stabilize and we'll move to a module and then, uh, hopefully. 528 00:29:43,169 --> 00:29:47,669 Like, uh, as that stabilizes further and the Linux community becomes used to 529 00:29:47,669 --> 00:29:49,050 us, then we would love to be mainlined. 530 00:29:49,080 --> 00:29:49,500 Even. 531 00:29:49,679 --> 00:29:52,110 Corey Quinn: I imagine you would have to open source a fair bit of that. 532 00:29:52,110 --> 00:29:55,860 It's like, well, except for this file system thing, that's a giant binary blo. 533 00:29:55,889 --> 00:29:55,979 Yeah. 534 00:29:55,979 --> 00:30:00,060 I, I see people having apoplectic fits already just at the notion. 535 00:30:00,090 --> 00:30:00,540 Hunter Leath: Yeah. 536 00:30:00,600 --> 00:30:03,360 And we'll, absolutely, like when we get to the point where it makes sense 537 00:30:03,360 --> 00:30:07,050 for us to be in the mainline kernel, we will absolutely open source the client. 538 00:30:07,824 --> 00:30:08,574 Corey Quinn: I look forward to it. 539 00:30:08,574 --> 00:30:12,235 I think this is going to be an interesting company to watch a d interesting 540 00:30:12,235 --> 00:30:15,655 space to watch, which again, I did not see myself saying at this late date. 541 00:30:15,685 --> 00:30:20,784 Hunter Leath: It is very interesting, I think, how much innovation is going 542 00:30:20,784 --> 00:30:23,844 to happen in the storage layer in the next 10 years, and I hope that 543 00:30:23,844 --> 00:30:25,465 we are an important part of that. 544 00:30:25,885 --> 00:30:31,675 But you'll see with time a lot of these AI workloads and the massive 545 00:30:31,675 --> 00:30:35,395 data centers of inference and training that open AI and Core Reef 546 00:30:35,395 --> 00:30:36,804 and Lambda Labs are building out. 547 00:30:37,110 --> 00:30:42,510 That people desperately need a faster, better way to get data into 548 00:30:42,510 --> 00:30:46,770 these GPUs, and so I expect that we will not be the only people in 549 00:30:46,770 --> 00:30:48,120 this space in the years to come. 550 00:30:49,170 --> 00:30:51,300 Corey Quinn: I really wanna thank you for taking the time to speak with me. 551 00:30:51,480 --> 00:30:54,150 If people wanna learn more, where's the best place for them to go? 552 00:30:54,210 --> 00:30:59,160 Hunter Leath: So the website is Archil.com, A-R-C-H-I-L, and then you can 553 00:30:59,160 --> 00:31:05,880 also find us on Twitter at Archil data or myself at JH Le L-E-H-T-H on Twitter, 554 00:31:06,360 --> 00:31:08,970 Corey Quinn: and we will of course put links to that in the show notes. 555 00:31:09,120 --> 00:31:11,460 Thank you so much for your time today, I appreciate it. 556 00:31:11,460 --> 00:31:12,210 Hunter Leath: Thank you, Corey. 557 00:31:12,470 --> 00:31:14,895 Hunter Leath, CEO at Archil. 558 00:31:14,895 --> 00:31:16,350 I'm Cloud economist 559 00:31:16,350 --> 00:31:18,879 Corey Quinn: Corey Quinn, and this is Screaming in the Cloud. 560 00:31:19,179 --> 00:31:22,659 If you've enjoyed this podcast, please, we have a five star review on your podcast 561 00:31:22,659 --> 00:31:26,560 platform of choice, whereas if you hated this podcast, please, we have a five star 562 00:31:26,560 --> 00:31:31,120 review on your podcast platform of choice, along with escaping comment, tearing 563 00:31:31,120 --> 00:31:35,470 apart the idea of this new file system without bothering to mention that your 564 00:31:35,470 --> 00:31:40,629 favorite file system was ripped out of the kernel after its creator killed his wife.