1 00:00:00,000 --> 00:00:04,590 You've found the backup wrap up your go-to podcast for all things 2 00:00:04,590 --> 00:00:07,200 backup recovery and cyber recovery. 3 00:00:07,770 --> 00:00:12,480 In this episode, we take a look at the use of artificial intelligence in backup. 4 00:00:12,930 --> 00:00:16,110 Can AI make your backup environment actually better? 5 00:00:16,810 --> 00:00:21,059 Prasanna Malaiyandi and I discuss AI and how it can help from 6 00:00:21,299 --> 00:00:25,409 possibly everything from scheduling backups to detecting ransomware. 7 00:00:25,919 --> 00:00:29,819 We talk about using it for deduplication, for capacity planning, 8 00:00:30,090 --> 00:00:33,389 and even helping you to write better disaster recovery plans. 9 00:00:33,989 --> 00:00:35,879 It's time to talk about AI and backups. 10 00:00:36,089 --> 00:00:36,899 Hope you enjoy it. 11 00:00:37,589 --> 00:00:41,609 By the way, if you don't know who I am, I'm w Curtis Preston, AKA, Mr. 12 00:00:41,609 --> 00:00:46,379 Backup, and I've been passionate about backup and recovery for over 30 years. 13 00:00:46,440 --> 00:00:50,405 Ever since I had to tell my boss I. That we had no backups of that really 14 00:00:50,405 --> 00:00:52,445 important database that we had just lost. 15 00:00:52,985 --> 00:00:56,315 I don't want that to happen to you, and that's why I do this podcast. 16 00:00:56,945 --> 00:01:01,925 On this podcast, we turn unappreciated backup admins into Cyber Recovery Heroes. 17 00:01:02,195 --> 00:01:04,025 This is the backup wrap up. 18 00:01:20,078 --> 00:01:21,128 Welcome to the show. 19 00:01:22,238 --> 00:01:27,158 Hi, I'm w Curtis Preston, AKA, Mr. Backup, and I have with me a guy who apparently 20 00:01:27,158 --> 00:01:29,168 doesn't know how to hold a coffee cup. 21 00:01:29,258 --> 00:01:31,238 Prasanna Malaiyandi, how's it going? 22 00:01:31,238 --> 00:01:31,758 Prasanna 23 00:01:31,943 --> 00:01:32,963 I am good, Curtis. 24 00:01:32,963 --> 00:01:36,143 I. So I think we need to clarify a 25 00:01:36,298 --> 00:01:37,508 are you defending yourself? 26 00:01:37,508 --> 00:01:39,338 Are you gonna try to defend your weirdness? 27 00:01:39,473 --> 00:01:41,573 I think we have to talk about multiple things. 28 00:01:42,323 --> 00:01:42,833 First. 29 00:01:42,833 --> 00:01:45,863 In India, they don't typically use like a mug. 30 00:01:46,253 --> 00:01:49,673 They use like a stainless steel cup, right? 31 00:01:49,733 --> 00:01:50,303 So if 32 00:01:50,303 --> 00:01:52,523 you're drinking hot beverages, you can only hold it from like the very, 33 00:01:52,523 --> 00:01:53,963 you saw it when we went to the Indian 34 00:01:53,963 --> 00:01:54,953 restaurant in San Diego, 35 00:01:55,058 --> 00:01:56,048 yeah, yeah. 36 00:01:56,183 --> 00:01:58,433 you have to hold it from the very top, otherwise you'll burn your hand. 37 00:01:59,468 --> 00:01:59,888 Right. 38 00:02:00,413 --> 00:02:03,323 And then most mugs, it just feels weird. 39 00:02:03,323 --> 00:02:06,563 Like I got, I got chunky fingers, like sausages, right? 40 00:02:06,623 --> 00:02:10,403 And so like putting it inside the mug, like the handle part of the mug. 41 00:02:10,823 --> 00:02:14,723 I feel like, especially if it's like a curve, not like a straight, I feel like 42 00:02:14,843 --> 00:02:16,403 there's not enough stability there. 43 00:02:17,603 --> 00:02:18,623 That's fascinating. 44 00:02:18,753 --> 00:02:20,493 So for people watching the video, who, by the 45 00:02:20,493 --> 00:02:23,283 way, we do publish a video on YouTube if you want to see our 46 00:02:23,283 --> 00:02:25,173 glorious faces and our expressions. 47 00:02:25,803 --> 00:02:29,343 But yeah, so when I hold a mug, I don't hold it like this through the 48 00:02:29,343 --> 00:02:29,913 handle. 49 00:02:29,913 --> 00:02:36,543 I basically grab it either from the top or I hold it like on the side. 50 00:02:36,543 --> 00:02:38,133 And then of course, the Pinky's kind 51 00:02:38,163 --> 00:02:38,838 The pinky, 52 00:02:39,123 --> 00:02:39,513 the bottom. 53 00:02:39,573 --> 00:02:42,453 But what's weird though is the pinky supporting the bottom thing. 54 00:02:43,218 --> 00:02:47,418 I know you've complained to me many times, but that's also how I hold my phone. 55 00:02:50,313 --> 00:02:51,423 you end up covering your microphone. 56 00:02:51,528 --> 00:02:52,818 always hold the phone and 57 00:02:52,818 --> 00:02:57,468 then my Pinky's kind of on the bottom, and so it always blocks the microphone. 58 00:02:58,248 --> 00:02:59,058 So Curtis is always 59 00:02:59,058 --> 00:03:00,738 like, were you underwater? 60 00:03:00,738 --> 00:03:01,908 Did you swallow your phone? 61 00:03:01,908 --> 00:03:02,958 What's going on? 62 00:03:03,723 --> 00:03:09,283 So regarding your defense from, you know, how they hold, do things in India. 63 00:03:09,463 --> 00:03:11,383 What part of India were you born in? 64 00:03:11,533 --> 00:03:12,313 Uh, just remind 65 00:03:12,433 --> 00:03:15,583 Yeah, I was, uh, born in not India, but, 66 00:03:19,453 --> 00:03:20,443 but at home. 67 00:03:20,803 --> 00:03:21,163 Right. 68 00:03:21,238 --> 00:03:21,658 But yeah, 69 00:03:21,658 --> 00:03:23,788 you were, you were raised by people born in 70 00:03:23,788 --> 00:03:26,398 India, and so you were, you were taught, 71 00:03:26,518 --> 00:03:27,028 yeah. 72 00:03:27,268 --> 00:03:30,448 And so actually I prefer, so even drinking water. 73 00:03:30,448 --> 00:03:32,068 I don't drink from a glass cup. 74 00:03:32,098 --> 00:03:33,568 I drink from a stainless steel cup. 75 00:03:34,168 --> 00:03:34,588 Right. 76 00:03:34,648 --> 00:03:37,888 Which is, if you haven't spent any time around, you know, 77 00:03:37,918 --> 00:03:39,838 Indians, you wouldn't know that. 78 00:03:39,838 --> 00:03:44,548 It's just that you use a lot, you use stainless steel for cups, for plates, 79 00:03:45,058 --> 00:03:45,148 right. 80 00:03:45,268 --> 00:03:48,268 As Curtis knows what I'm loading, the dishwasher and 81 00:03:48,268 --> 00:03:49,978 he's like, what is that racket? 82 00:03:50,338 --> 00:03:52,198 what is happening over there? 83 00:03:53,038 --> 00:03:54,628 Because everything's so noisy. 84 00:03:55,048 --> 00:03:58,588 They last longer and you don't have to worry about them breaking. 85 00:03:59,758 --> 00:04:01,798 That's, you know, I can't, I can't complain. 86 00:04:01,858 --> 00:04:02,068 Yeah. 87 00:04:03,298 --> 00:04:06,328 Uh, but yeah, I don't get the whole knot, you know? 88 00:04:06,508 --> 00:04:09,328 Here I am with four fingers in my mug. 89 00:04:09,658 --> 00:04:10,528 I'm just saying. 90 00:04:10,888 --> 00:04:14,128 Okay, so now what if that mug was smaller and the handle was curved, 91 00:04:14,623 --> 00:04:18,223 Well, then that's like a, that's like a girly mug and then, 92 00:04:18,253 --> 00:04:20,203 then you use two fingers like 93 00:04:20,203 --> 00:04:20,833 this. 94 00:04:21,088 --> 00:04:22,618 feel like it gives you enough stability? 95 00:04:23,833 --> 00:04:25,303 And yet I've never dropped a mug. 96 00:04:25,303 --> 00:04:25,453 I'm 97 00:04:25,453 --> 00:04:25,928 just saying. 98 00:04:26,158 --> 00:04:27,808 It's not from dropping the mug. 99 00:04:27,808 --> 00:04:29,308 It's from like when you, yeah. 100 00:04:29,308 --> 00:04:32,008 See when you're drinking it, it just feels like it's a little like, 101 00:04:33,158 --> 00:04:33,448 Yeah. 102 00:04:33,788 --> 00:04:34,008 Um, 103 00:04:34,318 --> 00:04:34,678 all over you. 104 00:04:34,968 --> 00:04:37,543 I just think you don't know how to hold a mic, but. 105 00:04:38,188 --> 00:04:41,368 Our listeners are probably like, what are these people talking about? 106 00:04:41,638 --> 00:04:44,128 By the way, this is a new format starting in the new year. 107 00:04:44,128 --> 00:04:46,828 We are now gonna just be talking about coffee and all the crazy 108 00:04:46,828 --> 00:04:47,968 things that Prasanna does. 109 00:04:49,378 --> 00:04:50,608 Yeah, absolutely. 110 00:04:51,178 --> 00:04:56,248 Um, or maybe we might actually talk about some stuff. 111 00:04:56,308 --> 00:05:03,058 So I thought, um, you know, we've been seeing, uh, AI on the news a lot, 112 00:05:03,118 --> 00:05:03,448 right? 113 00:05:03,658 --> 00:05:04,738 ai, I've never heard about it. 114 00:05:05,368 --> 00:05:07,168 Yeah, I've never, never heard of it. 115 00:05:07,168 --> 00:05:07,498 Yeah. 116 00:05:08,068 --> 00:05:13,708 So artificial intelligence, and if, if you've been following the backup 117 00:05:13,708 --> 00:05:22,708 industry much, you probably saw a few announcements from your, uh, backup 118 00:05:22,708 --> 00:05:27,688 company or maybe backup companies you're interested in about the use of ai. 119 00:05:28,168 --> 00:05:29,398 Within backup. 120 00:05:29,398 --> 00:05:31,858 And so I thought we'd talk about that a little bit, 121 00:05:31,948 --> 00:05:33,718 um, in this episode, and 122 00:05:33,778 --> 00:05:36,898 whether or not it has a use, right? 123 00:05:37,003 --> 00:05:43,093 And can, just to clarify, I think when a lot of these backup vendors launched ai, 124 00:05:43,603 --> 00:05:49,423 they were using AI for like the, not for the core product, right? 125 00:05:49,423 --> 00:05:54,733 So they were using AI for their support agent, or to help answer questions, right? 126 00:05:54,733 --> 00:05:58,393 Which I think we all understand, we all know about, but I think in this 127 00:05:58,393 --> 00:06:02,083 episode, I think we should focus on like the core part of backup. 128 00:06:03,148 --> 00:06:03,748 Yeah. 129 00:06:03,748 --> 00:06:06,628 So, so let's talk a just a little bit about, you know, 130 00:06:06,628 --> 00:06:08,398 what we mean when we say ai. 131 00:06:08,398 --> 00:06:12,238 There are different categories of ai and then also there's machine learning, which 132 00:06:12,238 --> 00:06:15,868 is very closely, and honestly, I, I, I, 133 00:06:16,408 --> 00:06:19,798 you know, I think I could describe the difference between machine 134 00:06:19,798 --> 00:06:22,588 learning and ai, but then there's something that, that. 135 00:06:22,963 --> 00:06:25,753 Changes, you know, that, that messes me up when we talk about that. 136 00:06:26,143 --> 00:06:29,773 Um, I'll just, for those of you that actually really know what AI 137 00:06:29,773 --> 00:06:32,773 is and machine learning is, you're gonna be offended by something 138 00:06:32,773 --> 00:06:34,153 I say during this episode. 139 00:06:34,153 --> 00:06:35,173 I, I'll just tell you that. 140 00:06:35,173 --> 00:06:38,593 But we're gonna use the terms almost interchangeably, but they're not. 141 00:06:38,953 --> 00:06:41,293 Uh, but I do want distinguish between. 142 00:06:41,698 --> 00:06:44,698 What is referred to as generative ai, right? 143 00:06:45,028 --> 00:06:49,138 Which is a, you know, a large language model that is 144 00:06:49,138 --> 00:06:52,198 going to create things there. 145 00:06:52,198 --> 00:06:53,668 It's not ex nihilo, right? 146 00:06:53,668 --> 00:06:55,378 It's not from, it's not from nothing. 147 00:06:55,378 --> 00:06:59,068 It's it, it has to, it has to have been trained on a large data set. 148 00:06:59,638 --> 00:07:01,998 But, those are the kinds of things that they're using, 149 00:07:01,998 --> 00:07:03,078 like you talked about there. 150 00:07:03,078 --> 00:07:04,068 Sup for support 151 00:07:04,068 --> 00:07:05,058 models, right? 152 00:07:05,208 --> 00:07:05,688 And, And, 153 00:07:05,688 --> 00:07:09,348 just as examples of large language models, you might've heard about 154 00:07:09,378 --> 00:07:15,528 meta's llama, lama three, Lama four, there's chat, GPT or open ais. 155 00:07:15,528 --> 00:07:16,098 What is it? 156 00:07:16,998 --> 00:07:18,078 OPT? 157 00:07:19,998 --> 00:07:20,298 What, 158 00:07:20,418 --> 00:07:21,438 the, the, actual model. 159 00:07:21,438 --> 00:07:21,528 the 160 00:07:21,528 --> 00:07:22,278 underlying model. 161 00:07:22,638 --> 00:07:23,358 Oh, okay. 162 00:07:23,363 --> 00:07:25,698 I, I, I would just, I would've just said chat, GPT. 163 00:07:25,728 --> 00:07:27,108 'cause everybody knows what chat GPT 164 00:07:27,108 --> 00:07:27,648 is, right? 165 00:07:27,948 --> 00:07:29,988 I mean, you've got copilot, you've got, you've 166 00:07:29,988 --> 00:07:33,688 got, Yeah, you, so you've got Claude from Anthropic. 167 00:07:33,778 --> 00:07:39,358 Um, there are a lot of people, you know, um, confused the company with the product. 168 00:07:39,358 --> 00:07:42,508 But, um, these are the, these are the ones that are grabbing 169 00:07:42,508 --> 00:07:43,588 all the headlines, right? 170 00:07:43,588 --> 00:07:47,668 They're also, they're also writing large bodies of texts. 171 00:07:47,668 --> 00:07:49,018 They're helping people to write books. 172 00:07:49,048 --> 00:07:51,508 They're helping people to do art. 173 00:07:51,568 --> 00:07:53,308 That, and there's a lot of, um. 174 00:07:54,283 --> 00:07:59,833 A lot of legal discussions around that, around the use of things like 175 00:07:59,833 --> 00:08:04,873 the books that I've written as, um, you know, feeding into that and, um, 176 00:08:05,863 --> 00:08:07,243 the, we're not talking about that, 177 00:08:07,903 --> 00:08:08,323 right? 178 00:08:08,653 --> 00:08:13,453 Um, we're not gonna talk about, Hey, um, chat GPT. 179 00:08:13,603 --> 00:08:14,863 My restore didn't work. 180 00:08:14,863 --> 00:08:16,693 Can you recreate all my documents? 181 00:08:17,263 --> 00:08:18,373 Um, it's not, 182 00:08:18,648 --> 00:08:21,373 there's not gonna be anything like that, at least not yet. 183 00:08:21,943 --> 00:08:28,933 Um, the, um, we're gonna talk about how AI can be used to basically 184 00:08:28,993 --> 00:08:31,603 enhance the core functionality. 185 00:08:31,603 --> 00:08:34,483 I mean, you said this in way, a fewer words a few minutes ago, 186 00:08:34,483 --> 00:08:40,363 but, uh, basically how it could be used to make backups better. 187 00:08:40,663 --> 00:08:44,743 And I think a good chunk of this is really, like you said, more 188 00:08:44,743 --> 00:08:46,573 around machine learning models, 189 00:08:47,233 --> 00:08:47,593 right, 190 00:08:47,803 --> 00:08:48,223 right, 191 00:08:48,403 --> 00:08:49,543 large language models. 192 00:08:50,203 --> 00:08:50,653 right. 193 00:08:50,653 --> 00:08:57,403 So the, the first section we will just talk about how potentially just talk about 194 00:08:58,243 --> 00:08:59,923 this is just sort of thoughts out loud. 195 00:08:59,923 --> 00:09:02,473 I know that we have a lot of vendors that listen to the podcast. 196 00:09:02,653 --> 00:09:02,953 We are. 197 00:09:03,508 --> 00:09:07,138 Technically aimed at the, the people who actually use backup and 198 00:09:07,138 --> 00:09:11,608 recovery, but I know a lot of vendors use the podcast, so feel free to 199 00:09:11,608 --> 00:09:13,408 take this episode and run with it and 200 00:09:13,408 --> 00:09:13,888 do stuff. 201 00:09:14,998 --> 00:09:20,248 So I, I guess the first question would be, do we think that, uh, machine learning 202 00:09:20,248 --> 00:09:26,278 can be used to help just to prove the efficiency of the backup process itself? 203 00:09:26,278 --> 00:09:26,938 What do you think about 204 00:09:27,133 --> 00:09:28,423 Oh, a thousand percent. 205 00:09:28,453 --> 00:09:29,863 A billion percent, Curtis. 206 00:09:30,223 --> 00:09:33,793 So I've never actually had to implement a backup system. 207 00:09:33,793 --> 00:09:34,333 But you've done 208 00:09:34,333 --> 00:09:35,263 tons of this, right? 209 00:09:35,263 --> 00:09:41,023 And how do you go about just planning your backup, right? 210 00:09:41,263 --> 00:09:42,973 How to back up an infrastructure, right? 211 00:09:42,973 --> 00:09:45,433 It's like, just walk us through that, right? 212 00:09:45,433 --> 00:09:49,633 And how many spreadsheets and all the rest that you have in 213 00:09:49,633 --> 00:09:51,013 order to try to optimize these. 214 00:09:51,733 --> 00:09:53,623 Yeah, I, I think about that a lot. 215 00:09:53,623 --> 00:09:58,183 And, and, and, and, and the answer is gonna depend greatly on the 216 00:09:58,183 --> 00:09:59,653 product that you're using, right? 217 00:09:59,683 --> 00:10:00,673 You know, I, I can think of. 218 00:10:01,228 --> 00:10:06,178 The traditional way is that you're going to create some kind of schedule, some 219 00:10:06,178 --> 00:10:08,908 kind of, uh, automatic backup schedule. 220 00:10:09,088 --> 00:10:13,558 Um, and you're going to do a, again, traditionally we'll 221 00:10:13,558 --> 00:10:14,608 do three categories here. 222 00:10:14,728 --> 00:10:17,458 Traditionally you've got some full backups and you're gonna do some 223 00:10:17,458 --> 00:10:19,348 full backups every once in a while. 224 00:10:19,678 --> 00:10:23,578 Um, and I was always a proponent if you had to do full backups, I was 225 00:10:23,578 --> 00:10:25,168 always a proponent of doing those. 226 00:10:25,648 --> 00:10:26,428 No. 227 00:10:27,088 --> 00:10:29,698 More often than once a month. 228 00:10:30,028 --> 00:10:34,018 Um, back in the days of tape, it was once a week because 229 00:10:34,813 --> 00:10:35,053 it, was 230 00:10:35,218 --> 00:10:36,898 complicated the restore process. 231 00:10:36,898 --> 00:10:37,198 Yeah. 232 00:10:37,768 --> 00:10:42,508 But, um, you know, doing it no more often than once a month, but depending on your 233 00:10:42,508 --> 00:10:44,218 backup product, you might be able to, to 234 00:10:44,218 --> 00:10:46,258 spread that out even over like three months. 235 00:10:46,258 --> 00:10:50,848 And then you also want to schedule, if your backup product 236 00:10:50,848 --> 00:10:54,088 is capable of doing it, you wanna schedule a cumulative incremental. 237 00:10:55,048 --> 00:10:56,758 A differential, some products call it. 238 00:10:57,418 --> 00:11:00,598 Um, and then of course the daily incremental. 239 00:11:00,598 --> 00:11:00,928 Right. 240 00:11:01,168 --> 00:11:02,278 So spreading 241 00:11:02,338 --> 00:11:02,938 that all 242 00:11:03,158 --> 00:11:05,623 for one application you're talking about, 243 00:11:05,913 --> 00:11:06,538 E exactly. 244 00:11:06,538 --> 00:11:09,778 You're doing this per application, per server. 245 00:11:10,348 --> 00:11:17,008 Um, and, and you're trying to load balance things out because if you've 246 00:11:17,008 --> 00:11:21,688 properly designed your system, it's probably not capable of doing a full 247 00:11:21,688 --> 00:11:23,638 backup of your environment in one night. 248 00:11:24,028 --> 00:11:24,418 Right. 249 00:11:24,748 --> 00:11:28,318 Um, because that would just be really expensive, and then the rest of the 250 00:11:28,678 --> 00:11:30,688 time it would go completely unused. 251 00:11:30,688 --> 00:11:31,108 Right? 252 00:11:31,738 --> 00:11:35,818 Um, so you, you buy it so that it's you, you size it so that it's big 253 00:11:35,818 --> 00:11:37,888 enough to do a full backup over time. 254 00:11:38,488 --> 00:11:45,388 And, um, you're right that, that, that scheduling that out is problematic, right? 255 00:11:45,808 --> 00:11:49,828 Um, and you, you definitely could use, um, uh, AI 256 00:11:49,828 --> 00:11:51,328 or ML to, to do that. 257 00:11:51,448 --> 00:11:53,158 And even for the scheduling aspect. 258 00:11:53,158 --> 00:11:56,428 So we talked about the applications, and then you were talking about sort 259 00:11:56,428 --> 00:11:59,758 of that infrastructure piece, which is shared and you now have to worry 260 00:11:59,758 --> 00:12:01,528 about it across all of these things. 261 00:12:02,038 --> 00:12:04,768 And I'm sure you had these bonkers spreadsheets that you 262 00:12:04,768 --> 00:12:06,808 were creating, trying to do this. 263 00:12:06,808 --> 00:12:08,968 Did it stretch all the way to the moon and back, by the way? 264 00:12:11,368 --> 00:12:15,988 Well, you know me for, it wasn't even a spreadsheet, it was just, uh, it, it was a 265 00:12:15,988 --> 00:12:16,558 script. 266 00:12:16,558 --> 00:12:16,828 Right. 267 00:12:16,828 --> 00:12:18,718 I would, I would just script all this nonsense. 268 00:12:18,718 --> 00:12:19,018 Right? 269 00:12:19,438 --> 00:12:22,378 Um, but it, but it, the bigger the environment, the more. 270 00:12:23,398 --> 00:12:26,488 That doing it programmatically made sense, right? 271 00:12:26,548 --> 00:12:30,928 Um, and, and by the way, even if you have a more modern backup tool 272 00:12:30,928 --> 00:12:35,218 that does incremental forever, there are many applications that 273 00:12:35,218 --> 00:12:36,658 won't, that won't let you do 274 00:12:36,658 --> 00:12:36,988 that. 275 00:12:37,018 --> 00:12:37,288 Right? 276 00:12:37,288 --> 00:12:41,188 I think of like database backups still need to be done every, you know, a full 277 00:12:41,188 --> 00:12:44,158 backup every so often, and you have to schedule these out, 278 00:12:44,473 --> 00:12:44,923 And that's the 279 00:12:44,923 --> 00:12:45,643 second category. 280 00:12:45,643 --> 00:12:47,443 'cause I know you talked about three categories. 281 00:12:48,478 --> 00:12:48,718 Yeah. 282 00:12:48,778 --> 00:12:49,228 Oh yeah. 283 00:12:49,228 --> 00:12:51,358 Oh, well the three categories were, yes. 284 00:12:51,388 --> 00:12:52,258 Uh, thank you. 285 00:12:53,188 --> 00:12:55,078 I'm glad I have you here sometimes, you know. 286 00:12:55,498 --> 00:12:55,738 Yeah. 287 00:12:55,738 --> 00:12:58,618 So you have the, the, the old school full and incremental, 288 00:12:58,618 --> 00:13:00,538 which old school is still current 289 00:13:00,538 --> 00:13:00,988 school. 290 00:13:00,988 --> 00:13:05,188 If we're talking about regular apps, then there's the forever incremental type. 291 00:13:05,613 --> 00:13:09,693 Um, and you don't, you, you do have to worry about scheduling those, 292 00:13:09,693 --> 00:13:12,933 but generally you just sort of tell 'em all to start at once and then 293 00:13:12,933 --> 00:13:17,523 they queue and then it is not, it's, it's a lot simpler to do those. 294 00:13:17,523 --> 00:13:22,858 I. But then the final category are ones that actually, um, and I 295 00:13:22,858 --> 00:13:26,608 think the one that probably stands out the most here would be Rubrik, 296 00:13:27,028 --> 00:13:27,328 right? 297 00:13:27,328 --> 00:13:30,658 Rubrik doesn't let you schedule, um, that 298 00:13:30,658 --> 00:13:31,048 stuff. 299 00:13:31,078 --> 00:13:33,238 You tell it what your RTO 300 00:13:33,238 --> 00:13:36,778 is and your RPO, and it just does the backups. 301 00:13:36,778 --> 00:13:40,468 I mean, in fact, there are people that complain that you cannot, at least 302 00:13:40,468 --> 00:13:42,358 last time I checked, you could not do. 303 00:13:42,803 --> 00:13:47,008 a a manually scheduled backup if you wanted to tell it when to do stuff. 304 00:13:47,428 --> 00:13:53,128 Um, I, I think this is probably the first use of some sort of machine learning 305 00:13:53,128 --> 00:13:56,518 or artificial intelligence that I can think of with regards to scheduling. 306 00:13:56,848 --> 00:13:58,858 Which, which I was also gonna chime in. 307 00:13:58,858 --> 00:14:01,588 So the first two methods you talked about, right? 308 00:14:01,858 --> 00:14:06,268 You're kind of statically doing this upfront, setting the schedules and 309 00:14:06,268 --> 00:14:09,358 hoping that forever that it will be good, 310 00:14:09,658 --> 00:14:09,958 Right. 311 00:14:09,958 --> 00:14:13,708 You'll always be able to meet it, but say that there's an additional load or a 312 00:14:13,708 --> 00:14:15,568 server goes down or something else, right. 313 00:14:15,568 --> 00:14:17,518 There's no way to fine tune and adjust that, 314 00:14:18,448 --> 00:14:21,718 Well, well, I, Well, there, I mean, there is, but there's 315 00:14:21,718 --> 00:14:23,293 no way to automatically fine 316 00:14:23,293 --> 00:14:24,093 tune and Yeah. 317 00:14:24,163 --> 00:14:24,453 Yeah. 318 00:14:24,658 --> 00:14:24,928 Right. 319 00:14:24,928 --> 00:14:28,228 And so you're just like, okay, maybe it'll fail a couple times 320 00:14:28,228 --> 00:14:31,648 and then I'll adjust the policies and then I'll be fine, but Right. 321 00:14:31,648 --> 00:14:35,458 Versus something like an SLA based, which I, I actually have 322 00:14:35,458 --> 00:14:36,868 looked at rubrics in the past, 323 00:14:36,898 --> 00:14:41,548 and I find that very enticing because really in the end, you 324 00:14:41,548 --> 00:14:43,168 care about what your RPO and RTO, 325 00:14:44,023 --> 00:14:44,203 Yeah. 326 00:14:44,203 --> 00:14:45,433 No one cares if you can back up. 327 00:14:45,433 --> 00:14:46,753 They only care if you can restore. 328 00:14:46,828 --> 00:14:52,438 the problem though is it's such a big paradigm shift for a lot of backup admins 329 00:14:53,098 --> 00:14:57,628 that it's very difficult to understand because it's like when people move 330 00:14:57,628 --> 00:15:00,388 from on-premises to the cloud and they were concerned because they're like, 331 00:15:00,418 --> 00:15:02,368 I can't touch and feel my equipment. 332 00:15:02,368 --> 00:15:02,758 Right. 333 00:15:03,058 --> 00:15:04,558 It's not something I could actually do. 334 00:15:04,558 --> 00:15:07,108 I think that's also the same challenges you get when you move 335 00:15:07,108 --> 00:15:11,938 from sort of, uh, schedule-based backups to sort of SLA based backups. 336 00:15:12,928 --> 00:15:15,418 Yeah, I, I liked, I liked the idea a lot. 337 00:15:15,583 --> 00:15:20,368 I, I, I still, again, you know, if I was, if I was running rubric, 338 00:15:20,368 --> 00:15:22,888 I would give people the ability to do a manual backup if they 339 00:15:22,888 --> 00:15:23,518 wanted to. 340 00:15:23,923 --> 00:15:27,418 But, but I do really like the idea of SLA driven backups, 341 00:15:27,598 --> 00:15:29,158 because I like the idea of SLAs. 342 00:15:29,158 --> 00:15:31,918 You know, we've talked about SLAs on here, and I like the idea of. 343 00:15:32,293 --> 00:15:36,373 Knowing the back backups were being done often enough to meet my SLAs. 344 00:15:36,373 --> 00:15:36,433 I 345 00:15:36,433 --> 00:15:37,483 really liked that idea. 346 00:15:38,113 --> 00:15:42,853 The one thing I think that is useful with these sort of approaches is 347 00:15:43,723 --> 00:15:46,873 we've talked about the fact that like your environment doesn't say static. 348 00:15:47,638 --> 00:15:47,988 Right. 349 00:15:48,568 --> 00:15:53,068 So as you're adding new workloads, as things are changing, you don't 350 00:15:53,068 --> 00:15:56,548 want to have to go recompute your entire spreadsheet or your 351 00:15:56,548 --> 00:15:58,108 script H every single time. 352 00:15:58,588 --> 00:16:03,328 So it's nice to have sort of these models that can automatically help fine tune and 353 00:16:03,328 --> 00:16:07,648 optimize so you're not wasting your time because it's more than likely that you're 354 00:16:07,648 --> 00:16:10,738 not gonna get it right the first time if you manually try to reset some of these 355 00:16:10,738 --> 00:16:11,008 things. 356 00:16:11,608 --> 00:16:14,548 And so having this automatic thing that constantly is 357 00:16:14,548 --> 00:16:16,948 adjusting just seems amazing. 358 00:16:18,118 --> 00:16:18,838 Yeah, it does. 359 00:16:18,838 --> 00:16:23,038 And I, and outside of Rubrik, I'm not aware of any tools that do that. 360 00:16:23,128 --> 00:16:25,498 Uh, but I, I think that this could certainly be a way where 361 00:16:25,498 --> 00:16:26,998 they could use AI to do that. 362 00:16:27,568 --> 00:16:29,068 Um, the. 363 00:16:30,073 --> 00:16:33,043 And I, and I was thinking about, again, going back to it, it's been a 364 00:16:33,043 --> 00:16:36,868 while since I've had to do this in a production environment, but the, the 365 00:16:36,873 --> 00:16:40,303 the first thing that you have to find out is how big is everything, right? 366 00:16:40,303 --> 00:16:43,603 How big is, is everything from a database perspective and 367 00:16:43,603 --> 00:16:45,073 how, how long does it take? 368 00:16:45,373 --> 00:16:47,923 'cause there's all these different, and that's the thing that nobody knows. 369 00:16:48,153 --> 00:16:48,333 Right. 370 00:16:48,333 --> 00:16:49,983 How big is your, how big is your data center? 371 00:16:50,133 --> 00:16:51,273 And they're like, I don't know. 372 00:16:51,543 --> 00:16:51,993 I don't know. 373 00:16:52,053 --> 00:16:54,393 And so like, you have to do a full backup first 374 00:16:54,393 --> 00:16:55,683 before you have any idea. 375 00:16:55,953 --> 00:16:59,253 And not every server backs up at the same speed and all these different things. 376 00:16:59,253 --> 00:17:00,543 So yeah, it it is a 377 00:17:00,543 --> 00:17:01,233 complicated 378 00:17:01,438 --> 00:17:03,538 and you may not be able to back up everything at the same 379 00:17:03,538 --> 00:17:04,408 time because there might be 380 00:17:04,408 --> 00:17:05,968 different hours, right? 381 00:17:05,968 --> 00:17:06,328 That 382 00:17:06,628 --> 00:17:11,248 a server is sort of offline or has less load that you can actually do it. 383 00:17:12,028 --> 00:17:17,638 Yeah, so having some sort of AI or ml, um, figure that out sounds amazing. 384 00:17:17,938 --> 00:17:18,328 Right? 385 00:17:18,658 --> 00:17:23,908 Another area where I think that this could help is very, very closely related, and 386 00:17:23,908 --> 00:17:29,398 that is, and, and some backup products do have this and that is making sure 387 00:17:29,398 --> 00:17:31,588 that everything in my data center. 388 00:17:31,963 --> 00:17:34,093 Is backed up in some 389 00:17:34,093 --> 00:17:35,113 way, right? 390 00:17:35,593 --> 00:17:40,033 Usually where you see this is an integration with like, um, uh, 391 00:17:40,063 --> 00:17:43,813 VMware or, uh, AWS, et cetera, right? 392 00:17:44,203 --> 00:17:49,903 Um, basically just connect to my entire, uh, you know, control 393 00:17:49,993 --> 00:17:53,893 panel and then just look and make sure that everything is connected 394 00:17:53,893 --> 00:17:56,053 to some type of policy to back it 395 00:17:56,053 --> 00:17:56,323 up. 396 00:17:56,668 --> 00:17:57,283 I, I think. 397 00:17:57,298 --> 00:18:00,358 a default policy if anything is created, so at least everything 398 00:18:00,358 --> 00:18:02,068 is protected, even though 399 00:18:02,068 --> 00:18:04,288 it may not be protected with the right thing, but at least it's 400 00:18:04,288 --> 00:18:06,298 being protected and you don't have to worry about these gaps. 401 00:18:06,298 --> 00:18:06,328 I. 402 00:18:07,003 --> 00:18:07,753 Exactly. 403 00:18:07,843 --> 00:18:08,623 Exactly. 404 00:18:08,743 --> 00:18:14,743 Um, and I, I think you do see this in a lot of backup products. 405 00:18:15,013 --> 00:18:17,233 Usually again, it's with integration 406 00:18:17,233 --> 00:18:22,933 with, uh, big things like VMware, HyperV, AWS, um, 407 00:18:22,993 --> 00:18:23,833 you know, et cetera. 408 00:18:24,283 --> 00:18:28,603 you need the companies, those vendors, to actually provide the APIs to be 409 00:18:28,603 --> 00:18:31,363 able to do these sort of queries, and I think that's where there's kind 410 00:18:31,363 --> 00:18:32,803 of a little bit of a tension there, 411 00:18:33,898 --> 00:18:34,408 Yeah. 412 00:18:35,038 --> 00:18:35,338 Yeah. 413 00:18:35,338 --> 00:18:39,478 I mean, theoretically you could scour the data center, right? 414 00:18:39,628 --> 00:18:41,308 Uh, looking for new computers. 415 00:18:41,728 --> 00:18:44,968 Again, I, I know I mentioned this before, but you know, back 416 00:18:44,968 --> 00:18:47,188 in the day we did that, right? 417 00:18:47,188 --> 00:18:49,168 And back in the day we did that with Vizio. 418 00:18:49,813 --> 00:18:52,483 Um, the, the vis, there used to be a very 419 00:18:52,483 --> 00:18:56,323 expensive version of Vizio that would just literally crawl your data center. 420 00:18:56,953 --> 00:19:00,163 And it used, uh, some very interesting technology. 421 00:19:00,613 --> 00:19:04,063 Um, I forgot the, the name of this, but like, inmap 422 00:19:04,303 --> 00:19:08,413 does this, where it, what it does is it sends a malformed packet. 423 00:19:09,253 --> 00:19:12,943 It finds an IP address, it sends a malformed packet to that IP address 424 00:19:12,943 --> 00:19:16,573 to see how it responds, and different things respond in different ways. 425 00:19:16,813 --> 00:19:18,673 And that's how it, that's how it, um, 426 00:19:18,943 --> 00:19:19,213 That 427 00:19:19,213 --> 00:19:20,653 is crazy that they built that. 428 00:19:21,613 --> 00:19:22,093 Yeah. 429 00:19:22,213 --> 00:19:22,633 Yeah. 430 00:19:22,753 --> 00:19:27,073 Um, and so you, you could theoretically do that, but a agreed, it's much easier 431 00:19:27,073 --> 00:19:32,353 if you just have, everything's gonna be in VMware or AWS and then just talk to AWS. 432 00:19:32,353 --> 00:19:36,853 Now again, going to VMware and AWS, there can be multiple virtual data centers. 433 00:19:37,093 --> 00:19:37,483 There can be 434 00:19:37,483 --> 00:19:39,073 multiple AWS accounts. 435 00:19:39,343 --> 00:19:42,823 So you, you, you want to make sure that, that you have some way to, to 436 00:19:42,823 --> 00:19:43,303 do that. 437 00:19:43,303 --> 00:19:44,803 And I, and I do like that idea. 438 00:19:45,658 --> 00:19:46,318 Shadow it. 439 00:19:47,083 --> 00:19:48,523 Yeah, shadow it bad, 440 00:19:48,583 --> 00:19:49,903 especially when it comes to backup. 441 00:19:49,963 --> 00:19:50,293 Right. 442 00:19:50,653 --> 00:19:56,443 Um, again, I'll tell a story from back in the day was the time that someone came to 443 00:19:56,443 --> 00:20:01,003 me and they had, they were DBAs and they, they gave me a directory of a database. 444 00:20:01,003 --> 00:20:02,053 They wanted me to restore. 445 00:20:02,083 --> 00:20:08,593 Restore, and it was temp, um slash TMP on a, on a HP box. 446 00:20:08,923 --> 00:20:16,243 And for those that don't know slash TMP on an HP box specifically, HPUX was in ram. 447 00:20:16,633 --> 00:20:19,633 So when you rebooted it, temp went away. 448 00:20:19,693 --> 00:20:21,073 And this, um, 449 00:20:21,133 --> 00:20:21,373 this 450 00:20:21,388 --> 00:20:22,228 it source code, 451 00:20:23,023 --> 00:20:23,563 what I. 452 00:20:23,578 --> 00:20:23,668 it? 453 00:20:23,668 --> 00:20:24,208 Source code 454 00:20:24,478 --> 00:20:25,498 It was source code. 455 00:20:25,768 --> 00:20:26,188 Yeah. 456 00:20:26,608 --> 00:20:29,728 And they were developing for months, like an entire team of 457 00:20:29,728 --> 00:20:33,778 developers developing source code of this new application in temp. 458 00:20:34,558 --> 00:20:39,718 And then we rebooted the server and they, and they came to me 459 00:20:39,718 --> 00:20:41,098 and asked me to restore it. 460 00:20:41,428 --> 00:20:44,788 And I was like, dude, we don't back up temp. I don't know 461 00:20:44,788 --> 00:20:45,658 what you're talking about. 462 00:20:45,688 --> 00:20:47,698 Like, and they're like, dude, this is really important, 463 00:20:47,698 --> 00:20:48,658 like heads are gonna roll. 464 00:20:48,658 --> 00:20:49,948 And I'm like, yeah, not mine. 465 00:20:50,578 --> 00:20:52,798 Like everybody knows we don't back up temp. 466 00:20:53,123 --> 00:20:54,833 Except for you, apparently. 467 00:20:55,438 --> 00:20:55,738 Oh 468 00:20:55,973 --> 00:20:59,213 Uh, so it's, I'm just, you know, it's really bad when you have 469 00:20:59,213 --> 00:21:02,843 a functioning system and then it's not being backed up again. 470 00:21:02,873 --> 00:21:07,823 Another story we used to have, um, we had a, a naming convention. 471 00:21:07,823 --> 00:21:09,083 Ours was very boring. 472 00:21:09,353 --> 00:21:13,343 Um, it, it was, it H-P-D-B-S-V-A, right? 473 00:21:13,343 --> 00:21:16,793 HP database server A, and there was HB FS oh one, et 474 00:21:16,793 --> 00:21:17,423 cetera, right? 475 00:21:18,628 --> 00:21:22,288 And I remember, and I had this form that you had to fill out. 476 00:21:22,378 --> 00:21:24,088 This was an actual piece of paper. 477 00:21:24,238 --> 00:21:24,508 We did 478 00:21:24,508 --> 00:21:25,798 not have web pages. 479 00:21:26,638 --> 00:21:26,788 Right? 480 00:21:27,418 --> 00:21:31,198 You had this form that you fill out and, and you had to, and, and it, it said 481 00:21:31,198 --> 00:21:35,818 on there, simply filling out this form is not, does not meet the requirement. 482 00:21:35,848 --> 00:21:40,498 You do not consider your system backed up until you have a signed form back from me. 483 00:21:40,738 --> 00:21:41,158 Right? 484 00:21:41,548 --> 00:21:44,188 And then one day somebody handed me a form and it said like. 485 00:21:44,578 --> 00:21:49,888 They wanted, like me to back up H-P-D-B-S-V-M, right? 486 00:21:50,158 --> 00:21:52,588 And I go, M that's interesting. 487 00:21:53,128 --> 00:21:58,528 The last server I remember hearing about was H. So that means there's an I, A 488 00:21:58,528 --> 00:22:01,168 J, A K, and an L out there somewhere. 489 00:22:01,228 --> 00:22:02,278 hasn't been backed up. 490 00:22:02,728 --> 00:22:04,048 That hasn't been backed up. 491 00:22:04,798 --> 00:22:05,428 Yeah. 492 00:22:05,938 --> 00:22:07,798 Um, so this idea of automatically 493 00:22:07,798 --> 00:22:10,738 detecting servers and applications sounds like a great 494 00:22:10,738 --> 00:22:11,248 idea. 495 00:22:11,368 --> 00:22:15,238 And also not just VMs, but also detect, it would be really 496 00:22:15,238 --> 00:22:16,648 nice if it detected the type of 497 00:22:16,648 --> 00:22:20,188 VM and said, this appears to be a SQL instance. 498 00:22:20,188 --> 00:22:21,928 We should back it up with the default SQL 499 00:22:21,928 --> 00:22:22,378 policy. 500 00:22:22,438 --> 00:22:23,068 That would be great. 501 00:22:24,013 --> 00:22:28,903 So in addition to making things more efficient, um, there are some 502 00:22:28,903 --> 00:22:32,983 other things we could do, uh, with AI that also would be interesting. 503 00:22:33,103 --> 00:22:34,093 Uh, what do 504 00:22:34,093 --> 00:22:35,298 you think is the, the first one? 505 00:22:35,398 --> 00:22:35,608 No. 506 00:22:35,638 --> 00:22:39,748 So I think one of the ones, and we've talked about it so much, so often, 507 00:22:39,808 --> 00:22:44,038 and vendors are starting to do this, it's around anomaly detection and 508 00:22:44,038 --> 00:22:46,738 it could be used in various fashion. 509 00:22:46,858 --> 00:22:52,798 So one thing is like, Hey, by the way, this server, all of a sudden it's backing 510 00:22:52,798 --> 00:22:55,348 up 10 times what it normally does. 511 00:22:55,348 --> 00:22:59,668 Maybe this might indicate like a malware or ransomware on the system. 512 00:23:00,388 --> 00:23:00,808 Right? 513 00:23:00,868 --> 00:23:01,348 Um. 514 00:23:01,608 --> 00:23:05,238 Or Hey, I've noticed that there's a bunch of data that's starting 515 00:23:05,238 --> 00:23:06,828 to look like based on entropy. 516 00:23:06,828 --> 00:23:09,828 That it's been encrypted, that doesn't look normal. 517 00:23:09,858 --> 00:23:12,708 Okay, maybe I should go investigate it, right? 518 00:23:12,708 --> 00:23:16,098 So, or it could even be security things like, Hey, you're logging 519 00:23:16,098 --> 00:23:19,638 in from a different place than normal as a backup admin. 520 00:23:19,638 --> 00:23:21,708 Is this the right thing or not? 521 00:23:22,423 --> 00:23:22,723 Yeah. 522 00:23:22,723 --> 00:23:28,673 And also very closely related to the stuff you said before was, uh, 523 00:23:28,943 --> 00:23:34,943 are files where the file type based on the first few bytes of the file, 524 00:23:35,303 --> 00:23:37,193 does not match the extension of the 525 00:23:37,193 --> 00:23:37,673 file. 526 00:23:37,913 --> 00:23:42,083 So it says it's a dot doc, but the first few bites of the file 527 00:23:42,083 --> 00:23:43,673 show that it's an application, for 528 00:23:43,748 --> 00:23:44,328 Sorry, one 529 00:23:44,328 --> 00:23:50,088 Yeah, that's an interesting use case around, uh, the first few bites because 530 00:23:50,088 --> 00:23:56,148 that could detect things that are being encrypted or other things that don't 531 00:23:56,148 --> 00:23:58,308 make sense, or potentially even malware. 532 00:23:58,308 --> 00:23:58,638 Right. 533 00:23:59,673 --> 00:24:03,813 Yeah, it, uh, it's something we do, you know, my, uh, employee is S two 534 00:24:03,813 --> 00:24:09,963 data and we do a lot of restores of old stuff, um, where we're pulling 535 00:24:09,963 --> 00:24:16,053 data off of tape often for, um, I. For e-discovery purposes and lawsuit 536 00:24:16,053 --> 00:24:19,023 purposes and, um, investigation purposes. 537 00:24:19,383 --> 00:24:22,953 And one of the things that we do as we're pulling data, 'cause we 538 00:24:22,953 --> 00:24:27,843 use a, a, a proprietary tool that we've written to restore data off 539 00:24:27,843 --> 00:24:33,243 of most backups rather than use the built in tool for a lot of reasons. 540 00:24:33,333 --> 00:24:37,683 Um, and this is one of them is that we check the file type against the file 541 00:24:37,683 --> 00:24:41,793 contents and, uh, it can, it can also indicate. 542 00:24:42,168 --> 00:24:44,538 Um, uh, subterfuge, 543 00:24:44,688 --> 00:24:45,108 right? 544 00:24:45,138 --> 00:24:47,478 Um, it can indicate somebody trying to hide something. 545 00:24:48,198 --> 00:24:51,768 Um, but yeah, so anomaly detection, I think is a really big one. 546 00:24:51,798 --> 00:24:52,608 Uh, right. 547 00:24:52,728 --> 00:24:58,908 Definitely that this is a, this is a, you looks like you've got ransomware, right? 548 00:24:58,908 --> 00:24:59,118 You need 549 00:24:59,118 --> 00:24:59,868 to solve that. 550 00:25:00,238 --> 00:25:02,818 That was probably the, the first big use of AI that I 551 00:25:02,818 --> 00:25:05,308 remember, uh, in, in the backup world. 552 00:25:05,728 --> 00:25:07,618 And I, I, I will say that if. 553 00:25:08,143 --> 00:25:11,983 The way that you know, that you have ransomware is that your backup 554 00:25:11,983 --> 00:25:16,213 product told you something is wrong, but, uh, but it, but it can 555 00:25:16,213 --> 00:25:16,633 happen. 556 00:25:16,663 --> 00:25:17,083 Right. 557 00:25:17,653 --> 00:25:23,443 Um, another one that I'll talk, uh, that I'd bring up is, is data classification. 558 00:25:23,563 --> 00:25:25,243 Again, I think that. 559 00:25:26,428 --> 00:25:29,878 This is, this is probably a very simple one, but the 560 00:25:29,878 --> 00:25:33,418 idea of like, looking at all the different data types and helping you to 561 00:25:33,478 --> 00:25:35,578 understand what is in your environment. 562 00:25:35,578 --> 00:25:37,258 This is not that new. 563 00:25:37,738 --> 00:25:43,308 Um, but perhaps the AI use case could be helping you to identify trends, 564 00:25:43,608 --> 00:25:47,328 um, and, and where the data's moving, where it's being created, where 565 00:25:47,328 --> 00:25:50,088 it's being changed, uh, et cetera. 566 00:25:50,608 --> 00:25:53,458 Um, and, and then, which is very closely related to my 567 00:25:53,458 --> 00:25:54,928 other idea, which is predictive 568 00:25:54,928 --> 00:25:55,678 analytics. 569 00:25:56,368 --> 00:25:56,788 Right. 570 00:25:56,818 --> 00:26:01,258 Um, again, going back to, uh, you know, back in the day, 571 00:26:01,528 --> 00:26:08,728 one of the things I remember being the hardest to do is capacity prediction. 572 00:26:08,728 --> 00:26:08,818 You 573 00:26:08,818 --> 00:26:12,458 know, predicting whether or not I have enough capacity To 574 00:26:12,458 --> 00:26:13,808 do my backups for the next six 575 00:26:13,873 --> 00:26:15,373 and you know what makes it even harder? 576 00:26:15,868 --> 00:26:16,348 What's that? 577 00:26:17,578 --> 00:26:20,848 It does, d ddu makes it way harder. 578 00:26:21,058 --> 00:26:24,268 And you know what AI right? 579 00:26:24,298 --> 00:26:30,808 Ai ml could, could use to, could be used because it's smarter than I am. 580 00:26:30,808 --> 00:26:31,948 Smarter than you are. 581 00:26:31,978 --> 00:26:34,498 It could actually understand the trends 582 00:26:34,498 --> 00:26:38,278 as to now what, what, let's talk about that Non, not every, 583 00:26:39,268 --> 00:26:40,798 everybody might not understand. 584 00:26:41,323 --> 00:26:44,443 Why DDU makes capacity, 585 00:26:44,563 --> 00:26:44,983 Sure. 586 00:26:45,253 --> 00:26:46,783 uh, management so 587 00:26:47,083 --> 00:26:50,173 So let's talk about the, before we get to D Dub, let's talk about like 588 00:26:50,173 --> 00:26:52,603 traditional storage or tape, right? 589 00:26:52,603 --> 00:26:53,023 So 590 00:26:53,473 --> 00:26:56,833 you're doing a full backup, you know how big your database is, therefore, 591 00:26:56,833 --> 00:27:00,433 you know, okay, my full backup is gonna take this much space and 592 00:27:00,433 --> 00:27:03,763 you know, with compression, maybe it's gonna be two x or half the space, right? 593 00:27:04,238 --> 00:27:08,258 And then, you know, okay, my daily change rate is say 5%, and based on the 594 00:27:08,258 --> 00:27:09,968 total size, I know what that's gonna be. 595 00:27:10,238 --> 00:27:10,598 And so 596 00:27:10,598 --> 00:27:14,018 if I'm doing weekly fulls, daily incrementals, I know how much 597 00:27:14,018 --> 00:27:15,398 storage I'm gonna need for a week. 598 00:27:16,468 --> 00:27:16,798 Yeah. 599 00:27:16,798 --> 00:27:20,758 And, and just as, and just as important, you also know how 600 00:27:20,758 --> 00:27:22,648 much storage, when you delete 601 00:27:23,218 --> 00:27:26,038 the, you know, the older backups. 602 00:27:26,038 --> 00:27:26,188 Yeah. 603 00:27:26,188 --> 00:27:29,818 You know how much storage will be freed up, which is just if, if not even more 604 00:27:29,818 --> 00:27:30,208 important. 605 00:27:30,763 --> 00:27:34,573 Now the problem with deduplication is they talk about these great rates like 606 00:27:34,573 --> 00:27:38,173 40 x, 30 x, 20 x, take your pick, right? 607 00:27:38,413 --> 00:27:39,433 And that's all great. 608 00:27:39,433 --> 00:27:44,143 If you're all like if a lot of your data is very similar, but it's hard 609 00:27:44,143 --> 00:27:48,013 to tell, is your data similar or not until you've actually start doing it. 610 00:27:48,013 --> 00:27:51,163 So if you're trying to buy storage for, say, three years 611 00:27:51,163 --> 00:27:52,813 ahead of time, a capacity plan. 612 00:27:53,578 --> 00:27:54,928 It becomes really difficult. 613 00:27:54,928 --> 00:27:56,188 And so you guess, right? 614 00:27:56,188 --> 00:27:58,588 You'll take a stab and maybe you look at some of your data and you're like, 615 00:27:58,588 --> 00:28:02,728 Hey, these kind of look the same, but you don't know if that's right or not 616 00:28:02,728 --> 00:28:04,558 until you actually start backing it up. 617 00:28:04,978 --> 00:28:09,238 And like you said, Curtis, if you go delete your backup, you may not 618 00:28:09,238 --> 00:28:12,508 actually free up that space because it's been de-duplicated against something 619 00:28:12,508 --> 00:28:13,798 else that you're still preserving. 620 00:28:14,623 --> 00:28:15,073 right, 621 00:28:15,133 --> 00:28:19,663 Say I go delete my backup for six months ago for one application. 622 00:28:19,873 --> 00:28:24,493 Another application might have, uh, common blocks with that data or with that other 623 00:28:24,493 --> 00:28:25,213 application. 624 00:28:25,453 --> 00:28:28,423 And so even though I deleted the first application's backup, 625 00:28:28,513 --> 00:28:29,743 it's not gonna free up space. 626 00:28:29,743 --> 00:28:33,253 And so you end up with this problem and this challenge. 627 00:28:33,313 --> 00:28:37,633 And that's one of the things, the hardest things about deduplication. 628 00:28:37,663 --> 00:28:41,143 Having worked at a company that did deduplication, customers 629 00:28:41,143 --> 00:28:42,013 always struggled with it, 630 00:28:43,338 --> 00:28:43,628 Yeah, 631 00:28:44,338 --> 00:28:44,998 And some of the 632 00:28:44,998 --> 00:28:47,848 things we would do is we would be like, Hey, let's scan your 633 00:28:47,968 --> 00:28:51,808 application and just understand what sort of DDU rates you may get. 634 00:28:52,138 --> 00:28:55,738 And even that's a guess, because maybe you move an application from one storage 635 00:28:55,738 --> 00:28:59,188 appliance to a different appliance and now your DDU rates are different. 636 00:29:00,808 --> 00:29:01,108 Yeah. 637 00:29:01,108 --> 00:29:02,188 And, and, and again, the 638 00:29:02,188 --> 00:29:05,938 one of the most frustrating things could be if you, you start. 639 00:29:06,808 --> 00:29:08,908 You're running outta capacity, right? 640 00:29:09,388 --> 00:29:13,168 And so you say, listen, I know we said we wanted to keep backups for 641 00:29:13,198 --> 00:29:16,588 three years, but we're running outta capacity and so we're gonna start 642 00:29:16,588 --> 00:29:18,838 deleting three years minus a month. 643 00:29:19,288 --> 00:29:20,698 And you do that and you get 644 00:29:20,698 --> 00:29:25,558 back 0.1% of your, it can be very difficult. 645 00:29:26,398 --> 00:29:27,148 Um, 646 00:29:27,223 --> 00:29:30,013 fact that to free up that space takes time. 647 00:29:30,013 --> 00:29:33,523 Because typically with a lot of these systems, there's a background process 648 00:29:33,523 --> 00:29:35,323 typically called garbage collection, 649 00:29:35,593 --> 00:29:40,003 which goes and now needs to free up all this data and that does take time to run. 650 00:29:40,648 --> 00:29:46,738 Yeah, it is, it is a two stage process where you, you, you, um, flag that 651 00:29:46,768 --> 00:29:49,348 block for deletion and then another 652 00:29:49,348 --> 00:29:52,168 process that runs typically when backups aren't running. 653 00:29:52,708 --> 00:29:56,608 Um, and you, you probably have to force the garbage collection process. 654 00:29:57,088 --> 00:29:59,278 Um, so go, go ahead. 655 00:29:59,638 --> 00:30:03,718 so I was just thinking as we were talking about the first time 656 00:30:03,718 --> 00:30:05,578 that I heard about AI in storage, 657 00:30:07,318 --> 00:30:11,908 and I think the first company that I can recall, and I'm sure there 658 00:30:11,908 --> 00:30:13,483 were others, was actually nimble. 659 00:30:14,398 --> 00:30:16,498 Storage and nimble. 660 00:30:16,498 --> 00:30:19,528 What they did is their first product when they built they, so 661 00:30:19,528 --> 00:30:20,788 they provided primary storage. 662 00:30:21,718 --> 00:30:27,178 And their first product, they basically were like, Hey, we are optimized for sql. 663 00:30:27,178 --> 00:30:29,158 We are optimized for VMware. 664 00:30:29,428 --> 00:30:32,098 We are optimized for these different, and I was like, oh, that's pretty awesome. 665 00:30:32,098 --> 00:30:33,328 They're doing it dynamically. 666 00:30:33,748 --> 00:30:36,898 But I think at the time it was kind of a static thing where you 667 00:30:36,898 --> 00:30:38,968 would say, Hey, I have VMware. 668 00:30:38,968 --> 00:30:40,888 I'm writing into this data store. 669 00:30:41,428 --> 00:30:45,208 And it would optimize its, and it would basically pick different 670 00:30:45,208 --> 00:30:46,918 block sizes for deduplication 671 00:30:47,113 --> 00:30:49,033 Right, right, right. 672 00:30:49,513 --> 00:30:49,813 Yeah. 673 00:30:49,813 --> 00:30:50,563 That's interesting. 674 00:30:50,648 --> 00:30:56,893 The, the, the, I, I, I think div, going back to the thing 675 00:30:56,893 --> 00:30:59,563 we were talking about of like. 676 00:30:59,968 --> 00:31:05,038 Using AI to basically help me understand when do I need to order more storage? 677 00:31:05,428 --> 00:31:07,408 It can, to the best of its ability. 678 00:31:07,588 --> 00:31:11,428 It can actually look at all of the DDU rates, right? 679 00:31:11,428 --> 00:31:13,738 At all of the at, at what? 680 00:31:13,768 --> 00:31:17,008 It could look at the DDU rate of each individual backup, right? 681 00:31:17,008 --> 00:31:19,498 You, you gave, you told me it's a backup this much and this is 682 00:31:19,498 --> 00:31:20,908 how much, and so we can actually 683 00:31:20,968 --> 00:31:23,938 run all those calculations and I can actually figure out. 684 00:31:24,258 --> 00:31:26,958 Well in six months, based on if everything stays the 685 00:31:26,958 --> 00:31:29,238 same in six months, you're gonna be 686 00:31:29,238 --> 00:31:29,838 outta storage. 687 00:31:29,838 --> 00:31:29,988 So 688 00:31:30,073 --> 00:31:31,513 many vendors actually do. 689 00:31:32,478 --> 00:31:32,808 Yeah. 690 00:31:32,868 --> 00:31:33,228 Yeah. 691 00:31:33,978 --> 00:31:36,528 Um, so the, the, um, 692 00:31:37,603 --> 00:31:39,943 Because I think storage capacity is a little easier. 693 00:31:40,528 --> 00:31:44,368 To predict, because like you said, you're not really changing things, right. 694 00:31:44,368 --> 00:31:45,928 You know what your policy is. 695 00:31:45,928 --> 00:31:49,468 You know what data's coming in, you know how long it's, you're keeping it, 696 00:31:49,468 --> 00:31:53,038 you know what your deduplication rates are, you know how much it's filling up. 697 00:31:53,038 --> 00:31:57,208 So I think it's a little easier than what we had talked about previously 698 00:31:57,208 --> 00:32:00,148 where it's like, okay, now let me plan out my entire backup infrastructure 699 00:32:00,148 --> 00:32:01,318 and start scheduling that. 700 00:32:01,873 --> 00:32:02,263 Yeah. 701 00:32:02,263 --> 00:32:07,003 Speaking of dedupe, can AI help dedupe itself? 702 00:32:07,003 --> 00:32:07,663 Do you think that? 703 00:32:08,743 --> 00:32:09,223 can. 704 00:32:10,428 --> 00:32:12,843 So I think my biggest. 705 00:32:12,843 --> 00:32:18,543 Challenge would be that to run AI requires compute 706 00:32:19,323 --> 00:32:20,643 and usually backup. 707 00:32:20,643 --> 00:32:22,293 You want to go as fast as you can, 708 00:32:23,103 --> 00:32:23,523 Mm-hmm. 709 00:32:23,913 --> 00:32:24,213 right? 710 00:32:24,213 --> 00:32:25,893 And so I think there's that tension. 711 00:32:26,958 --> 00:32:31,728 That exists between running as fast as you can versus introducing 712 00:32:31,728 --> 00:32:35,388 something in the pipeline to that could potentially slow things down. 713 00:32:35,778 --> 00:32:39,378 And you'd have to also ask at what cost, right? 714 00:32:39,378 --> 00:32:44,028 Like, are you going to be saving, say 70% additional versus a traditional 715 00:32:44,028 --> 00:32:46,998 algorithms, or is it gonna be much less 716 00:32:48,288 --> 00:32:57,383 Yeah, I think ddu in, um, in the backup world, there, there, there 717 00:32:57,383 --> 00:33:02,213 have been two main ways to do ddu, which has been, there has been 718 00:33:02,753 --> 00:33:04,673 something that isn't really ddu, but 719 00:33:05,003 --> 00:33:08,873 there were DDU products that called themselves DDU products that did this. 720 00:33:09,563 --> 00:33:12,923 Uh, and that would be block level, um, 721 00:33:12,983 --> 00:33:14,273 incremental, essentially. 722 00:33:14,303 --> 00:33:14,603 Right? 723 00:33:14,603 --> 00:33:14,843 Not 724 00:33:14,843 --> 00:33:16,883 actually de-duping things against each other, but just. 725 00:33:17,243 --> 00:33:22,973 Using technology to lower the additional new data that's 726 00:33:22,973 --> 00:33:24,083 backed up from each workload. 727 00:33:24,428 --> 00:33:27,758 But then the traditional ddu, the way it works for those that don't know 728 00:33:27,758 --> 00:33:31,658 this, is that you slice it up, you slice everything up into what are 729 00:33:31,658 --> 00:33:33,398 typically called shards or chunks. 730 00:33:33,908 --> 00:33:37,628 You run some type of algorithm on it that gives you some type of thing. 731 00:33:37,628 --> 00:33:38,228 Like, like 732 00:33:38,663 --> 00:33:39,263 A fingerprint. 733 00:33:39,338 --> 00:33:41,048 the original SHA two, 734 00:33:41,048 --> 00:33:42,128 SHA 2 56. 735 00:33:42,128 --> 00:33:47,318 And again, here the, the better the algorithm, um, the better the ddu, 736 00:33:47,348 --> 00:33:50,848 but the better the algorithm, the more compute it takes going back to 737 00:33:50,848 --> 00:33:51,928 your trade off thing. 738 00:33:52,438 --> 00:33:57,358 And so, um, that's the way basically every chunk it's run through, you come 739 00:33:57,358 --> 00:34:01,288 up with this alpha numeric string, that alpha numeric string is compared 740 00:34:01,288 --> 00:34:03,058 with every other alpha numeric string. 741 00:34:03,058 --> 00:34:07,168 I. Um, and then that's how you identify redundant data. 742 00:34:07,228 --> 00:34:11,368 And one of the challenges you have with that method is that, uh, the data slides, 743 00:34:11,968 --> 00:34:16,768 um, and so if you don't slice the data at exactly the same spot it, it's duplicate 744 00:34:16,768 --> 00:34:18,688 data, but you don't, don't identify it. 745 00:34:19,228 --> 00:34:24,028 The, there is a completely different way which, um, you 746 00:34:24,028 --> 00:34:26,248 look at the way vast does things. 747 00:34:26,503 --> 00:34:28,573 They do something completely different, right? 748 00:34:28,573 --> 00:34:31,663 So they, they have an algorithm and, and I, I'm guessing they 749 00:34:31,663 --> 00:34:34,333 use AI or ML to, to, do this. 750 00:34:34,333 --> 00:34:41,203 They have an algorithm that, um, basically identifies data that 751 00:34:41,203 --> 00:34:43,723 is probably redundant, right? 752 00:34:43,753 --> 00:34:48,133 Um, that, that, so they, they've got two different ways to do de-dupe and I, so 753 00:34:48,133 --> 00:34:51,433 there are potentially, again, potentially. 754 00:34:53,308 --> 00:35:00,268 AI or ML could be used to identify a new way to identify duplicate 755 00:35:00,268 --> 00:35:02,458 data that is maybe, maybe 756 00:35:02,458 --> 00:35:06,358 more efficient from a compute and storage. 757 00:35:06,688 --> 00:35:09,808 Like even if it was just more efficient from a compute standpoint, 758 00:35:09,808 --> 00:35:13,528 but got the but got the same amount of dedupe, that would still 759 00:35:13,528 --> 00:35:14,218 be great. 760 00:35:14,908 --> 00:35:16,828 Um, but 761 00:35:16,858 --> 00:35:18,448 potentially this is something 762 00:35:18,448 --> 00:35:19,708 that I think, uh, AI could 763 00:35:19,858 --> 00:35:24,088 and the one thing I did also want to comment on Curtis is, uh, going back to 764 00:35:24,088 --> 00:35:27,538 your comment about, okay, if the data shifts, then now you have to make sure 765 00:35:27,538 --> 00:35:30,028 that you're doing the right blocks, right? 766 00:35:30,238 --> 00:35:34,078 Uh, this is where companies though have done sort of, uh, what you're 767 00:35:34,078 --> 00:35:35,578 talking about is called fixed block. 768 00:35:35,758 --> 00:35:36,988 Fixed block deduplication, 769 00:35:37,558 --> 00:35:37,918 right? 770 00:35:38,188 --> 00:35:38,428 There are 771 00:35:38,428 --> 00:35:40,948 many vendors out there though, who do variable size. 772 00:35:41,743 --> 00:35:46,243 Variable block, uh, deduplication, which allows it to vary such that if 773 00:35:46,243 --> 00:35:51,133 you do get an offset right, because of some data change, it's still able to 774 00:35:51,133 --> 00:35:53,473 dup everything else after that because 775 00:35:53,473 --> 00:35:57,913 of how it's actually computing the chunks, the segments, right? 776 00:35:58,153 --> 00:35:58,513 Each of 777 00:35:58,513 --> 00:35:59,083 the blocks. 778 00:35:59,818 --> 00:36:00,208 Yep. 779 00:36:00,808 --> 00:36:06,688 Um, so, uh, so that, that's certainly an area where, where AI could potentially 780 00:36:06,688 --> 00:36:12,658 help the, um, the next, do you think it could help with recovery testing? 781 00:36:13,093 --> 00:36:15,253 Oh yeah, I would. 782 00:36:15,943 --> 00:36:20,293 So one thing for C is like, most people probably don't 783 00:36:20,293 --> 00:36:22,003 know how to write a DR plan, 784 00:36:22,688 --> 00:36:23,108 Mm-hmm. 785 00:36:23,188 --> 00:36:23,588 Mm-hmm. 786 00:36:24,253 --> 00:36:24,733 right. 787 00:36:25,003 --> 00:36:31,243 Um, I wonder if you took ai, like even, and I'm going back to the first 788 00:36:31,243 --> 00:36:32,773 set, right, the large language models, 789 00:36:33,643 --> 00:36:34,123 Yep. 790 00:36:34,633 --> 00:36:35,623 So the thing we said we 791 00:36:35,623 --> 00:36:37,813 weren't talking about, I think we're gonna talk about it here. 792 00:36:38,008 --> 00:36:38,338 Yeah. 793 00:36:38,518 --> 00:36:42,058 I think at least to start with, it's like, Hey, here's all my data. 794 00:36:42,058 --> 00:36:43,618 Here's my applications. 795 00:36:43,858 --> 00:36:45,928 Help me build a DR test plan. 796 00:36:46,783 --> 00:36:47,173 Yeah, 797 00:36:47,503 --> 00:36:48,433 I like that idea. 798 00:36:48,868 --> 00:36:49,168 And 799 00:36:49,168 --> 00:36:52,198 see what it pops out because, and it may not be perfect, and don't just 800 00:36:52,198 --> 00:36:55,798 blindly trust what it provides, but use it as a starting point, right? 801 00:36:55,798 --> 00:36:56,968 And then go use that. 802 00:36:56,968 --> 00:36:59,248 Because I think a lot of people struggle with, where do I even start? 803 00:37:00,553 --> 00:37:01,213 Yeah. 804 00:37:01,513 --> 00:37:06,103 And you could also, um, you could use it like a chaos monkey, 805 00:37:06,133 --> 00:37:06,523 right? 806 00:37:06,523 --> 00:37:07,153 You could use it. 807 00:37:07,243 --> 00:37:09,583 Help me come up with some interesting scenarios. 808 00:37:09,898 --> 00:37:14,128 To just make the, the idea, you know, one of the things that we talked about with in 809 00:37:14,128 --> 00:37:18,148 terms of, uh, cyber testing, uh, was, um. 810 00:37:18,658 --> 00:37:21,478 You know, when we had Mike on the idea of like, doing this and, and 811 00:37:21,478 --> 00:37:25,318 making it, making it fun, making it a game, uh, I like that idea a 812 00:37:25,318 --> 00:37:27,028 lot and I think maybe AI could help 813 00:37:27,028 --> 00:37:27,478 there. 814 00:37:27,988 --> 00:37:28,588 Um, 815 00:37:28,648 --> 00:37:33,658 if, if it helps you do recovery testing more often, um, and, uh, 816 00:37:33,688 --> 00:37:37,888 helps you identify potential, uh, uh, plot, I was gonna say plot 817 00:37:37,888 --> 00:37:43,498 holes, uh, potential, potential holes in your program, uh, then that, then that 818 00:37:43,498 --> 00:37:45,268 I think could be, um, very 819 00:37:45,268 --> 00:37:45,568 helpful. 820 00:37:45,763 --> 00:37:50,503 And Curtis, since you threw out a term, Chaos Monkey is a tool that was released 821 00:37:50,503 --> 00:37:55,843 by Netflix, and literally what it is used for is to just test it, resiliency. 822 00:37:56,173 --> 00:37:59,773 So it'll go randomly, kill services, kill locations, kill 823 00:37:59,863 --> 00:38:02,203 network connections, just to see. 824 00:38:02,848 --> 00:38:06,748 Is streaming, interrupted, are, uh, end users having any sort of 825 00:38:06,748 --> 00:38:11,098 issues and it's able to do this at a scale and in an automated fashion 826 00:38:11,098 --> 00:38:13,978 versus someone like trying to think about all the combinations, 827 00:38:13,978 --> 00:38:17,098 permutations, and scenarios, because they're probably gonna miss things. 828 00:38:17,098 --> 00:38:20,548 And so Netflix designed this thing to actually go out and 829 00:38:20,548 --> 00:38:21,718 test their infrastructure. 830 00:38:22,828 --> 00:38:24,208 It is pretty impressive. 831 00:38:24,268 --> 00:38:27,208 Uh, you know, their infrastructure in general is pretty impressive. 832 00:38:27,208 --> 00:38:28,738 It's not flawless. 833 00:38:28,858 --> 00:38:32,128 Um, I did, I did watch part of the, uh. 834 00:38:32,833 --> 00:38:36,883 The Tyson fight a little while ago, and that was on Netflix 835 00:38:36,883 --> 00:38:39,193 and it was not good, right? 836 00:38:39,343 --> 00:38:42,283 That wasn't so much a resilient thing as it was. 837 00:38:42,283 --> 00:38:45,613 They just, again, they could have used perhaps a little bit better 838 00:38:45,613 --> 00:38:49,483 AI to predict the, what kind of load they were gonna have. 839 00:38:49,993 --> 00:38:50,803 But yeah. 840 00:38:50,863 --> 00:38:56,053 But the idea of predicting crazy things that will happen, uh, Netflix 841 00:38:56,053 --> 00:38:59,653 is pretty darn resilient, uh, when it comes to their infrastructure, 842 00:38:59,713 --> 00:38:59,893 Yep. 843 00:39:00,043 --> 00:39:02,053 yeah, I, I like that idea a lot. 844 00:39:02,113 --> 00:39:05,443 Um, and, and I think, I think this is something that could be, that, 845 00:39:05,448 --> 00:39:09,073 that, that, again, an, uh, uh, an LLM could actually help with, right? 846 00:39:09,073 --> 00:39:11,953 So, like I said, the thing that we said we weren't gonna talk about, 847 00:39:11,953 --> 00:39:12,913 we could talk about it, right? 848 00:39:13,333 --> 00:39:17,623 Um, and for those, if you've never used a chat, g PT or a Claude, 849 00:39:17,983 --> 00:39:19,693 uh, I think it's very useful 850 00:39:19,693 --> 00:39:20,383 here, right? 851 00:39:20,383 --> 00:39:23,023 You, you could say, Hey, I, I'm this kind of company. 852 00:39:23,848 --> 00:39:26,188 This is the type of company, you know, and I understand the, 853 00:39:26,188 --> 00:39:27,628 the privacy concerns of what you 854 00:39:27,628 --> 00:39:29,338 share with a chat g pt or a clot. 855 00:39:29,638 --> 00:39:32,878 Uh, there, there are, by the way, there are on-prem versions that 856 00:39:32,878 --> 00:39:36,748 you can run, uh, of these LLMs too, so that you can keep the 857 00:39:36,748 --> 00:39:37,918 data to yourself. 858 00:39:38,038 --> 00:39:41,698 But the, you have a conversation with it. 859 00:39:41,698 --> 00:39:45,328 Here's the type of company I am, here's the type of computing environment I have. 860 00:39:45,598 --> 00:39:46,678 What do you th what could go 861 00:39:46,678 --> 00:39:47,188 wrong? 862 00:39:47,638 --> 00:39:51,178 Um, you know what, what could I build a, a dr scenario 863 00:39:51,178 --> 00:39:51,448 around? 864 00:39:51,788 --> 00:39:52,958 Any final thoughts? 865 00:39:52,958 --> 00:39:57,008 Can you think of, uh, any other areas where we could use AI and, and backup? 866 00:39:59,003 --> 00:39:59,843 Not so much. 867 00:39:59,843 --> 00:40:05,303 I think the one thing I do wanna call out though is AI is here to stay. 868 00:40:05,363 --> 00:40:06,563 ML is here to stay. 869 00:40:06,593 --> 00:40:07,703 Don't be afraid of it. 870 00:40:08,003 --> 00:40:08,573 Use it. 871 00:40:09,443 --> 00:40:13,583 Right in the right ways and don't be afraid and just start thinking about it. 872 00:40:14,033 --> 00:40:19,433 Uh, the one other thing I will call out is as companies are starting 873 00:40:19,433 --> 00:40:24,863 to dig into AI and ML for their own applications, production applications 874 00:40:24,863 --> 00:40:28,313 and other things, as a backup admin, you need to start thinking 875 00:40:28,313 --> 00:40:30,563 about how do I protect this, right? 876 00:40:30,563 --> 00:40:31,433 How do I back it up? 877 00:40:31,433 --> 00:40:32,783 How would I potentially restore it? 878 00:40:32,783 --> 00:40:36,053 Because there's a lot of data and training these models. 879 00:40:36,653 --> 00:40:38,603 Is really, really expensive. 880 00:40:39,368 --> 00:40:39,588 Mm. 881 00:40:39,713 --> 00:40:44,243 And so you wanna make sure you have mechanisms to protect the models 882 00:40:44,243 --> 00:40:49,258 that emerge from all of this training so you can restore them if needed. 883 00:40:50,363 --> 00:40:55,013 So use backup to, to make AI more resilient while AI makes backup more 884 00:40:55,013 --> 00:40:55,553 resilient. 885 00:40:56,418 --> 00:40:56,778 I like that. 886 00:40:57,383 --> 00:40:59,333 We'll call that a symbiosis. 887 00:40:59,423 --> 00:41:00,233 I like that a lot. 888 00:41:01,103 --> 00:41:05,633 Uh, one my final thought is that potentially you could use, again, 889 00:41:05,633 --> 00:41:08,034 going back to the thing we said we weren't gonna talk about. 890 00:41:08,708 --> 00:41:12,848 You could use LLMs to help select vendors, right? 891 00:41:12,878 --> 00:41:15,668 You could say, Hey, here are all my requirements and here's all the 892 00:41:15,668 --> 00:41:21,368 documents that they, they gave me this 57 page response to my 10 page RFI. 893 00:41:21,368 --> 00:41:22,838 Can you help me make sense of it? 894 00:41:23,198 --> 00:41:26,978 Um, and, uh, you, you could use that again, trust but 895 00:41:26,978 --> 00:41:28,988 verify when using an LLM for 896 00:41:28,988 --> 00:41:29,348 sure. 897 00:41:30,188 --> 00:41:33,968 All right, well, thanks again, Prasanna, uh, for a good chat. 898 00:41:34,208 --> 00:41:35,258 Thank you, Curtis. 899 00:41:35,258 --> 00:41:38,198 And I am not gonna change how I hold a coffee mug. 900 00:41:38,198 --> 00:41:38,618 I'm sorry. 901 00:41:40,718 --> 00:41:42,338 I, I would expect no less. 902 00:41:42,698 --> 00:41:45,668 And thanks to our listeners, uh, we'd be nothing without you. 903 00:41:45,818 --> 00:41:46,803 That is a wrap.