1 00:00:35,062 --> 00:00:38,032 W. Curtis Preston: Hi, and welcome to Backup Central's Restored All podcast. 2 00:00:38,152 --> 00:00:40,792 I'm your host, Debbie Curtis Preston, a k a, Mr. 3 00:00:40,797 --> 00:00:45,202 Backup, and I with me, a guy that I'm hoping knows a little 4 00:00:45,202 --> 00:00:46,042 something about plumbing. 5 00:00:46,652 --> 00:00:47,172 Prasanna. 6 00:00:47,432 --> 00:00:48,712 Malaiyandi, how's it going? 7 00:00:49,537 --> 00:00:50,587 Prasanna Malaiyandi: I am good, Curtis. 8 00:00:50,797 --> 00:00:51,157 W. Curtis Preston: how's that? 9 00:00:51,157 --> 00:00:51,607 YouTube? 10 00:00:51,667 --> 00:00:51,937 How's that? 11 00:00:51,942 --> 00:00:55,237 YouTube, the, the YouTube knowledge that you have, does it cover plumbing 12 00:00:56,617 --> 00:00:57,757 Prasanna Malaiyandi: kind of, sort of. 13 00:00:57,757 --> 00:01:03,847 Unfortunately it's not something I'm as familiar with, but depends 14 00:01:03,852 --> 00:01:04,807 on what you're looking to do. 15 00:01:05,767 --> 00:01:06,127 W. Curtis Preston: Yeah. 16 00:01:06,132 --> 00:01:07,987 I gotta replace tonight. 17 00:01:08,077 --> 00:01:09,697 I gotta replace a couple of valves. 18 00:01:09,847 --> 00:01:11,827 Um, Shut off valves. 19 00:01:12,067 --> 00:01:16,927 Um, one of them, I, one of them is on the, the, it's the one that goes 20 00:01:16,927 --> 00:01:22,537 behind my fridge and in order to like, it's like I, I can't easily, 21 00:01:22,687 --> 00:01:25,357 it's a, it's a, what do you call it? 22 00:01:25,357 --> 00:01:29,197 It's the, the threaded, no, it's not a compression fit. 23 00:01:29,197 --> 00:01:30,727 It's the, it's a threaded pipe. 24 00:01:32,767 --> 00:01:37,897 Like, uh, I really gotta get in there and get a hold of the, the upper part 25 00:01:37,897 --> 00:01:42,877 of the pipe, uh, in order to hold that pipe still while I'm undoing it. 26 00:01:42,877 --> 00:01:46,897 Cuz it, my first attempts at it did not, I don't want to be 27 00:01:46,897 --> 00:01:48,717 cracking off a pipe up in the wall. 28 00:01:49,072 --> 00:01:51,232 Prasanna Malaiyandi: Yeah, I was just gonna say that's, yeah, 29 00:01:51,232 --> 00:01:52,372 the last thing you want to do. 30 00:01:52,642 --> 00:02:00,472 I was thinking for the shutoff though, I'm assuming you took the handle off 31 00:02:02,482 --> 00:02:03,082 W. Curtis Preston: The handle 32 00:02:03,922 --> 00:02:05,062 Prasanna Malaiyandi: to give you more space 33 00:02:06,682 --> 00:02:07,042 W. Curtis Preston: it? 34 00:02:07,372 --> 00:02:08,212 No, there's no handle. 35 00:02:08,212 --> 00:02:10,132 It's just, I mean, the handle's like integrated. 36 00:02:10,132 --> 00:02:10,882 It's really small. 37 00:02:11,722 --> 00:02:12,862 It's not a space problem. 38 00:02:13,132 --> 00:02:17,452 It's just, uh, it's um, you know, 39 00:02:18,082 --> 00:02:19,252 Prasanna Malaiyandi: Or they also say heat. 40 00:02:20,532 --> 00:02:21,532 W. Curtis Preston: Yeah, give it 41 00:02:21,562 --> 00:02:24,892 Prasanna Malaiyandi: So when I saw my, oh, so when we had our water heater 42 00:02:24,892 --> 00:02:27,292 replaced, I was watching cuz I'm always curious to see what people do. 43 00:02:27,772 --> 00:02:29,872 And our pipes were really, really bad at old. 44 00:02:30,467 --> 00:02:33,382 because the last time our water heater was replaced was like 15 years ago. 45 00:02:34,012 --> 00:02:37,612 And so the guy's like, you just heat it up really, really hot and then 46 00:02:37,612 --> 00:02:41,962 you tighten it first to break the stuff off and then you loosen it. 47 00:02:42,637 --> 00:02:43,267 W. Curtis Preston: Oh, interesting. 48 00:02:43,477 --> 00:02:43,717 Yeah. 49 00:02:43,717 --> 00:02:44,497 Maybe I'll see. 50 00:02:44,587 --> 00:02:45,547 I'll take a look at that. 51 00:02:45,922 --> 00:02:46,312 Prasanna Malaiyandi: Yeah. 52 00:02:47,047 --> 00:02:47,527 W. Curtis Preston: Uh, 53 00:02:47,632 --> 00:02:50,302 Prasanna Malaiyandi: took like in a settling torch or a butane torch 54 00:02:50,302 --> 00:02:53,602 and then heated up nice and hot, and then he tightened it just a little. 55 00:02:53,647 --> 00:02:54,877 W. Curtis Preston: what, I'm not gonna be able to do that. 56 00:02:54,877 --> 00:02:55,357 You know why? 57 00:02:55,537 --> 00:02:55,927 There's this 58 00:02:55,942 --> 00:02:57,022 Prasanna Malaiyandi: Oh, the drywall. 59 00:02:57,532 --> 00:02:57,712 Oh, 60 00:02:58,387 --> 00:02:59,857 W. Curtis Preston: Plastic, worse than dry. 61 00:03:00,837 --> 00:03:01,012 Prasanna Malaiyandi: yeah. 62 00:03:02,137 --> 00:03:07,177 W. Curtis Preston: Um, so hey, we're gonna talk about, uh, we're, you 63 00:03:07,177 --> 00:03:12,997 know, in our continuing, um, backup to basic series, you know, we talked 64 00:03:12,997 --> 00:03:17,167 about, like recently we talked about a difference between backup and archive. 65 00:03:17,217 --> 00:03:20,457 , in a previous episode we talked about protecting your backup data. 66 00:03:21,267 --> 00:03:31,197 This one, um, you know, this one is, um, it's about backup levels, right? 67 00:03:31,737 --> 00:03:35,787 And, you know, it's, honestly, I'll just say this isn't the most 68 00:03:36,187 --> 00:03:44,562 exciting topic, Um, and maybe, maybe, we'll, maybe we'll make it a short 69 00:03:44,562 --> 00:03:46,572 episode if it's not interesting. 70 00:03:46,632 --> 00:03:49,392 You know, we're not gonna make, we're not gonna make a long episode cuz 71 00:03:49,392 --> 00:03:53,892 this isn't gonna be one that's gonna be, uh, littered with stuff about 72 00:03:53,892 --> 00:03:55,812 cyber attacks and things like that. 73 00:03:55,812 --> 00:03:56,652 Backup levels 74 00:03:56,737 --> 00:03:58,177 Prasanna Malaiyandi: really straightforward. 75 00:03:59,162 --> 00:04:00,087 W. Curtis Preston: really straightforward. 76 00:04:00,087 --> 00:04:00,477 Yeah. 77 00:04:00,897 --> 00:04:04,527 Uh, I do want to help for people that sometimes struggle with 78 00:04:04,527 --> 00:04:06,297 the different types of backups. 79 00:04:06,447 --> 00:04:08,727 Um, I want them to help them understand. 80 00:04:09,657 --> 00:04:13,277 I, I'd say that, you know, if you just go and I'm, uh, throughout 81 00:04:13,277 --> 00:04:16,307 our usual disclaimer Prasanna, and I work for different companies. 82 00:04:16,307 --> 00:04:21,047 I work for Druva, he works for Zoom, and the, this is an independent podcast. 83 00:04:21,047 --> 00:04:22,067 It's not theirs. 84 00:04:22,067 --> 00:04:24,497 And it's, uh, the opinions that you hear are ours. 85 00:04:25,097 --> 00:04:29,567 Uh, be sure to rate us, uh, go to your favorite, um, pod catcher and 86 00:04:29,567 --> 00:04:33,017 uh, scroll down to wherever they got the stars and give us all the stars. 87 00:04:33,347 --> 00:04:36,497 And the best thing is if you give us a comment, those are always nice, 88 00:04:36,857 --> 00:04:38,837 especially on Apple Podcasts, which is. 89 00:04:39,762 --> 00:04:42,812 the primary way that people listen to us over here. 90 00:04:43,322 --> 00:04:49,802 And, um, also if you wanna join the conversation, I am w Curtis Preston gmail 91 00:04:50,192 --> 00:04:54,302 and also WC Preston for now on Twitter. 92 00:04:54,572 --> 00:04:55,052 I dunno what's 93 00:04:55,197 --> 00:04:55,697 Prasanna Malaiyandi: do you have a master on? 94 00:04:55,862 --> 00:04:58,652 W. Curtis Preston: Twitter man, I don't have a Macedon account. 95 00:04:58,657 --> 00:05:00,302 Uh, I am interested in that. 96 00:05:00,482 --> 00:05:04,262 We'll see if, I don't know if the average person can, can 97 00:05:04,357 --> 00:05:05,087 Prasanna Malaiyandi: I don't think so. 98 00:05:05,087 --> 00:05:06,377 From what everything I've heard. 99 00:05:06,497 --> 00:05:06,887 Yeah. 100 00:05:08,207 --> 00:05:09,677 W. Curtis Preston: It took me a minute, right? 101 00:05:09,737 --> 00:05:11,957 Twitter, you know, you sign up for account, you're done right here. 102 00:05:11,957 --> 00:05:14,207 It's like, well, what community do you want it attach to? 103 00:05:14,267 --> 00:05:16,067 I'm like, oh, it's so complicated. 104 00:05:16,067 --> 00:05:18,047 I don't wanna know what community, you know? 105 00:05:18,347 --> 00:05:20,117 Um, we'll see. 106 00:05:20,597 --> 00:05:24,317 But, uh, anyway, at WC Preston on Twitter, maybe I should sign it 107 00:05:24,317 --> 00:05:26,807 for my, at WC Preston on Mastodon. 108 00:05:27,287 --> 00:05:28,217 Make sure I get that. 109 00:05:28,652 --> 00:05:30,602 Prasanna Malaiyandi: I think you were throwing out the disclaimer though, 110 00:05:30,602 --> 00:05:34,712 because for this chapter, if people want to actually read the chapter and 111 00:05:34,712 --> 00:05:36,812 understand all these terms, right. 112 00:05:36,812 --> 00:05:39,512 Druva is offering a free ebook version. 113 00:05:41,357 --> 00:05:43,247 W. Curtis Preston: I, you know, that wasn't why I was saying 114 00:05:43,247 --> 00:05:44,777 it, but, we'll, we'll do that. 115 00:05:44,957 --> 00:05:48,047 So we are discussing my book, modern Data Protection. 116 00:05:48,047 --> 00:05:51,857 There's a picture of it with the little armadillo for those of you watching on 117 00:05:51,857 --> 00:05:54,407 the video version on backup central.com. 118 00:05:55,187 --> 00:06:01,337 And uh, you can get a free ebook copy of it by going to druva.com/ebook. 119 00:06:01,337 --> 00:06:04,517 That's d r uva.com/ebook. 120 00:06:05,447 --> 00:06:05,987 And. 121 00:06:07,202 --> 00:06:09,407 while supplies last, or while the contract lasts. 122 00:06:09,407 --> 00:06:13,157 So, you know, this, this, this episode will live for a long time. 123 00:06:13,157 --> 00:06:15,827 If, if you go there and it's not there, it's not my fault. 124 00:06:16,577 --> 00:06:20,927 So, uh, first off, you know, again, this is pretty basic, but. 125 00:06:22,052 --> 00:06:24,152 You know, you should know what a full backup is. 126 00:06:24,152 --> 00:06:24,632 Obviously. 127 00:06:24,632 --> 00:06:25,572 That's it. 128 00:06:25,577 --> 00:06:26,882 It backs up everything, right? 129 00:06:27,392 --> 00:06:28,322 Typically. 130 00:06:28,322 --> 00:06:29,222 Well, not typically. 131 00:06:29,342 --> 00:06:31,622 You always have to do a full backup before you can do any, 132 00:06:31,622 --> 00:06:33,032 any kind of incremental backup. 133 00:06:33,032 --> 00:06:34,622 That's the, that's the next type. 134 00:06:35,252 --> 00:06:40,892 The, there are amazingly, a handful of different types of 135 00:06:40,892 --> 00:06:42,272 the traditional incremental 136 00:06:42,482 --> 00:06:44,432 Prasanna Malaiyandi: but before you get to incremental, so. 137 00:06:44,642 --> 00:06:45,092 W. Curtis Preston: Yeah. 138 00:06:45,692 --> 00:06:49,902 Prasanna Malaiyandi: Even for fulls, are there different types of fulls? 139 00:06:53,012 --> 00:06:57,302 W. Curtis Preston: Well, there's not different levels for fulls, but there 140 00:06:57,302 --> 00:06:59,852 are different ways that fulls get done. 141 00:07:01,352 --> 00:07:01,952 Right. 142 00:07:01,982 --> 00:07:06,332 Like you, like we could be doing a file system backup or we could be doing a block 143 00:07:06,332 --> 00:07:11,402 level backup, but either way you're still backing up all the things, all the bites. 144 00:07:11,462 --> 00:07:11,852 Yeah. 145 00:07:12,152 --> 00:07:18,752 Um, the, uh, and it's, it's the full is the thing that puts the most amount of, 146 00:07:19,112 --> 00:07:22,052 of stress on the thing being backed up. 147 00:07:22,112 --> 00:07:22,502 Right. 148 00:07:23,192 --> 00:07:26,212 The, this is why we try to limit them whenever. 149 00:07:27,602 --> 00:07:35,192 , and this is why, this is why backups broke when VMware took off because everybody was 150 00:07:35,192 --> 00:07:37,862 just doing full and incremental backups and they weren't coordinating them. 151 00:07:37,862 --> 00:07:40,072 And you know, it was just you, 152 00:07:40,277 --> 00:07:41,657 Prasanna Malaiyandi: lot of resources being used. 153 00:07:42,092 --> 00:07:43,592 W. Curtis Preston: Yeah, exactly right. 154 00:07:43,622 --> 00:07:48,932 And um, if you do, if you do a nightly full backup, or let's say the weekend 155 00:07:48,932 --> 00:07:54,032 full backup, and you do 'em all on Friday, And you've got 20 different 156 00:07:54,472 --> 00:07:56,432 physical servers, no big deal. 157 00:07:56,552 --> 00:08:00,722 But if you do 20 full backups on 20 VMs that are all on one 158 00:08:01,112 --> 00:08:03,317 physical server, this is a problem. 159 00:08:03,407 --> 00:08:03,767 Right. 160 00:08:04,607 --> 00:08:07,787 Um, so I've got a couple different types here. 161 00:08:07,787 --> 00:08:11,567 There's the, the typical incremental backup, which is, it basically, 162 00:08:12,167 --> 00:08:16,037 it backs up everything since the last backup of whatever kind. 163 00:08:16,817 --> 00:08:21,107 And typically what what you do, and this is with backups that have 164 00:08:21,107 --> 00:08:24,917 levels, you do a full backup, then you do a series of incremental 165 00:08:24,917 --> 00:08:27,317 backups, and then you do another full. 166 00:08:28,217 --> 00:08:28,382 right? 167 00:08:28,382 --> 00:08:30,512 That's the, that's the sort of the typical way. 168 00:08:30,962 --> 00:08:33,512 There is something called a cumulative incremental backup. 169 00:08:34,172 --> 00:08:37,412 Um, I prefer that to the term differential backup, and I can 170 00:08:37,412 --> 00:08:38,432 talk about that in a minute. 171 00:08:39,092 --> 00:08:44,732 But a cumulative incremental backup is essentially just that it backs up 172 00:08:44,732 --> 00:08:51,152 everything that has changed since the last full different, um, different product. 173 00:08:51,857 --> 00:08:53,027 Uh, do different things. 174 00:08:53,057 --> 00:08:58,517 Uh, the differential backup is a term that you see a lot in windows and, um, 175 00:08:59,707 --> 00:09:07,667 I, I, and I, I don't use it because I find the term, um, the, the, because 176 00:09:10,217 --> 00:09:12,767 depending on where you ask the term differential means different things. 177 00:09:12,767 --> 00:09:13,757 And so I, I don't like it. 178 00:09:13,997 --> 00:09:14,267 Right? 179 00:09:14,267 --> 00:09:16,277 So I like the term cumulative incremental. 180 00:09:17,072 --> 00:09:20,702 Even if your backup product doesn't use it, just learn 181 00:09:20,702 --> 00:09:22,922 what, whatever they call it. 182 00:09:22,982 --> 00:09:23,312 What? 183 00:09:23,687 --> 00:09:24,137 Prasanna Malaiyandi: wait. 184 00:09:24,647 --> 00:09:30,917 So as far as I understand it, , you have the thing that's like full then. 185 00:09:31,007 --> 00:09:37,577 So say I do a full on the weekend, right on Monday, I do, uh, incremental, 186 00:09:37,982 --> 00:09:38,542 W. Curtis Preston: Uhhuh. 187 00:09:38,837 --> 00:09:39,077 Prasanna Malaiyandi: right? 188 00:09:39,082 --> 00:09:41,837 Which will be the differences for whatever happened between Sunday and Monday, 189 00:09:42,512 --> 00:09:43,002 W. Curtis Preston: Yeah. 190 00:09:43,067 --> 00:09:43,397 Prasanna Malaiyandi: right? 191 00:09:43,607 --> 00:09:45,617 Tuesday I do another incremental 192 00:09:46,362 --> 00:09:46,517 W. Curtis Preston: Yeah, 193 00:09:47,117 --> 00:09:47,387 Prasanna Malaiyandi: right? 194 00:09:47,387 --> 00:09:49,607 That'll be the differences between Monday and Tuesday. 195 00:09:50,537 --> 00:09:50,927 W. Curtis Preston: yeah, 196 00:09:51,932 --> 00:09:53,792 Prasanna Malaiyandi: And then Wednesday I do another incremental. 197 00:09:53,792 --> 00:09:55,202 That's the difference between Tuesday and Wednesday. 198 00:09:55,742 --> 00:09:59,792 Now, for different, uh, for cumulative incrementals, it's, I 199 00:09:59,792 --> 00:10:03,602 did my still, I'll do my Sunday, full Monday I'll do my cumulative. 200 00:10:04,052 --> 00:10:06,782 Incremental, or yeah, incremental, which will just be 201 00:10:07,292 --> 00:10:08,942 Sunday to Monday, differences. 202 00:10:09,002 --> 00:10:11,612 Tuesday, when I do the cumulative incremental, it'll 203 00:10:11,872 --> 00:10:12,962 be from Sunday to Tuesday 204 00:10:13,757 --> 00:10:14,297 W. Curtis Preston: Correct. 205 00:10:14,672 --> 00:10:17,252 Prasanna Malaiyandi: and Wednesday when I do a cumulative incremental, it'll 206 00:10:17,252 --> 00:10:18,602 be the differences between Sunday and. 207 00:10:19,937 --> 00:10:20,537 W. Curtis Preston: Correct. 208 00:10:21,467 --> 00:10:26,027 And so what, what we used to do back in the day was I would do a monthly 209 00:10:26,027 --> 00:10:30,437 full weekly cumulative incrementals and daily incrementals, right? 210 00:10:30,437 --> 00:10:34,697 So that way, even if I'm 28 days into the month, I would only need 211 00:10:34,697 --> 00:10:36,137 three backups to get the job done. 212 00:10:37,062 --> 00:10:37,247 . Right. 213 00:10:37,877 --> 00:10:41,267 Um, I, I don't use the term differential. 214 00:10:41,297 --> 00:10:45,107 Differential is what a lot of places call accumulative a criminal backup. 215 00:10:45,112 --> 00:10:49,037 But again, different products use it to mean different things, so I don't use it. 216 00:10:49,487 --> 00:10:53,657 Um, but interestingly enough, what we haven't talked about is backup 217 00:10:53,657 --> 00:10:59,897 levels, which is the term that, again, you don't see too much these days. 218 00:11:00,497 --> 00:11:01,817 Have you, have you, are you seeing it? 219 00:11:03,077 --> 00:11:06,237 Prasanna Malaiyandi: Very, very rarely do Does anyone talk. 220 00:11:06,587 --> 00:11:10,397 The only place I've seen levels really is when you're talking 221 00:11:10,397 --> 00:11:13,937 like database backups, but most. 222 00:11:13,997 --> 00:11:14,387 W. Curtis Preston: what? 223 00:11:14,392 --> 00:11:15,017 That is true. 224 00:11:15,227 --> 00:11:15,587 Right? 225 00:11:15,712 --> 00:11:15,767 Prasanna Malaiyandi: Yeah. 226 00:11:15,887 --> 00:11:21,137 But most backups software itself no longer talks about the levels. 227 00:11:21,137 --> 00:11:21,347 Right. 228 00:11:21,347 --> 00:11:24,137 It's either fulls or incrementals. 229 00:11:24,347 --> 00:11:25,757 W. Curtis Preston: I think it was too complicated. 230 00:11:26,327 --> 00:11:26,657 Yeah. 231 00:11:26,657 --> 00:11:29,117 The basically if to, to do. 232 00:11:29,867 --> 00:11:34,007 To do a full on a weekend and then daily regular incrementals, you would 233 00:11:34,007 --> 00:11:38,387 do a 0, 1, 2, 3, 4, 5, 6, 7, right? 234 00:11:38,387 --> 00:11:39,227 Well, seven. 235 00:11:39,232 --> 00:11:45,077 And then maybe, um, so then, and again, this is, this is impossible to do without 236 00:11:45,077 --> 00:11:53,027 a whiteboard, but if you did, um, a zero followed by all the way up to the five, if 237 00:11:53,277 --> 00:11:56,507 you did a one again following that five. 238 00:11:57,667 --> 00:11:59,987 you would get all the changes since to zero. 239 00:12:00,542 --> 00:12:00,932 Prasanna Malaiyandi: Hmm. 240 00:12:01,427 --> 00:12:02,477 W. Curtis Preston: So it was accumulative. 241 00:12:02,477 --> 00:12:03,527 Incremental, right. 242 00:12:04,037 --> 00:12:08,147 But what some people would do is they would do this, um, this, 243 00:12:08,567 --> 00:12:11,387 there, there's, there was this thing called towers of Hanoi, which, 244 00:12:12,107 --> 00:12:14,957 um, it, it comes from the game. 245 00:12:15,182 --> 00:12:15,302 Prasanna Malaiyandi: like 246 00:12:16,232 --> 00:12:16,682 W. Curtis Preston: Yes. 247 00:12:16,682 --> 00:12:17,012 Yes. 248 00:12:17,012 --> 00:12:18,152 It's like the thing with the game. 249 00:12:18,152 --> 00:12:18,482 Yes, 250 00:12:18,622 --> 00:12:18,822 Prasanna Malaiyandi: three. 251 00:12:18,822 --> 00:12:19,222 Yeah. 252 00:12:19,622 --> 00:12:20,732 W. Curtis Preston: Comes from that game. 253 00:12:21,242 --> 00:12:27,482 Um, and so the, and again, this is really, don't try to understand this, I'm just 254 00:12:27,542 --> 00:12:32,072 putting it out there just for, this is for the true backup anoracks, so who's gonna 255 00:12:32,192 --> 00:12:33,842 Prasanna Malaiyandi: Daniel Rosen Hill. 256 00:12:33,842 --> 00:12:34,382 We're looking. 257 00:12:34,922 --> 00:12:35,792 Daniel Rose Hill. 258 00:12:35,792 --> 00:12:36,872 We're looking at you. 259 00:12:37,862 --> 00:12:38,642 W. Curtis Preston: Exactly. 260 00:12:38,972 --> 00:12:41,912 So, um, let's see. 261 00:12:42,422 --> 00:12:51,062 So what I've got here is, , um, a towers of Illinois Schedule 0 3 2 5 4 7 6. 262 00:12:52,022 --> 00:12:56,882 And what it does is it, is it like every file that's changed 263 00:12:56,882 --> 00:12:59,432 ends up on two different, um, 264 00:12:59,462 --> 00:12:59,822 Prasanna Malaiyandi: Mm, 265 00:13:00,872 --> 00:13:02,372 W. Curtis Preston: two different, um, 266 00:13:02,552 --> 00:13:03,002 Prasanna Malaiyandi: days. 267 00:13:04,112 --> 00:13:04,802 W. Curtis Preston: backups. 268 00:13:04,922 --> 00:13:05,312 Prasanna Malaiyandi: Yeah, 269 00:13:05,882 --> 00:13:08,072 W. Curtis Preston: Um, and um, 270 00:13:08,072 --> 00:13:10,802 Prasanna Malaiyandi: you don't run the risk of if a backup is bad, 271 00:13:10,802 --> 00:13:15,182 you've lost the file, which could happen if you just had a full and you 272 00:13:15,182 --> 00:13:16,712 never did another incremental again. 273 00:13:18,047 --> 00:13:18,557 W. Curtis Preston: Yeah. 274 00:13:18,797 --> 00:13:22,127 Prasanna Malaiyandi: Or sorry you never did another full again, I, I 275 00:13:22,132 --> 00:13:28,577 think one thing we forgot to mention Curtis though, is the benefits of doing 276 00:13:28,582 --> 00:13:32,357 a a, we might have talked about it implicitly, but the benefits of doing a 277 00:13:32,807 --> 00:13:36,407 cumulative incremental versus a normal incremental when it comes to restores. 278 00:13:36,947 --> 00:13:39,017 W. Curtis Preston: Well, you know, here's the thing. 279 00:13:39,947 --> 00:13:43,457 The benefits were much bigger when we were using tape. 280 00:13:44,607 --> 00:13:44,687 Prasanna Malaiyandi: Hmm. 281 00:13:46,012 --> 00:13:49,232 W. Curtis Preston: I, I think that as, and, and what we're gonna discover is 282 00:13:49,232 --> 00:13:54,782 that this is why a lot of this has, I think, gone by the wayside is because, 283 00:13:55,502 --> 00:14:01,292 because when we were using tape, you'd gr you'd grab the full tape, you would grab 284 00:14:01,297 --> 00:14:05,072 the cumulative incremental tape, and you would grab any incremental since then. 285 00:14:05,372 --> 00:14:10,742 So you've got maximum like eight tapes if you did a full, 286 00:14:10,742 --> 00:14:12,302 followed by 30 incremental. 287 00:14:13,307 --> 00:14:18,857 , you would, um, you would need 28 tapes, you know, at the end of the month. 288 00:14:19,247 --> 00:14:21,737 So this way you only need like a handful of tapes. 289 00:14:22,397 --> 00:14:25,397 And what it allowed you to do was it allowed you to do 290 00:14:25,907 --> 00:14:27,709 less frequent full backups. 291 00:14:29,483 --> 00:14:31,883 Because it was a, it was a, an extra cost. 292 00:14:31,883 --> 00:14:34,313 You're like, well, by doing a weekly cumulative incremental, 293 00:14:34,313 --> 00:14:37,403 I could do a monthly full backup instead of a weekly full backup. 294 00:14:37,793 --> 00:14:41,543 Thus really reducing how big the backup system needed to be, 295 00:14:42,023 --> 00:14:43,793 how many tapes I needed to use. 296 00:14:44,453 --> 00:14:48,563 Um, And, and by doing that cumulative incremental, it made 297 00:14:48,563 --> 00:14:49,973 the restore much, much faster. 298 00:14:50,813 --> 00:14:54,563 I would argue if you have a decent backup system today and you've got a monthly 299 00:14:54,563 --> 00:14:59,543 full and 30 days of incremental, or even 60 days of incremental, the difference 300 00:14:59,543 --> 00:15:05,273 between restore times, um, depending on how you store the backup data, shouldn't 301 00:15:05,273 --> 00:15:10,133 be as significant as they were back in the day when we were loading tapes, cuz 302 00:15:10,133 --> 00:15:12,113 every tape loaded was like two minutes 303 00:15:12,508 --> 00:15:12,623 Prasanna Malaiyandi: Yeah. 304 00:15:12,923 --> 00:15:13,523 But, 305 00:15:13,583 --> 00:15:14,213 W. Curtis Preston: to the first. 306 00:15:14,903 --> 00:15:19,343 Prasanna Malaiyandi: but I do wonder though if when you are doing 307 00:15:19,343 --> 00:15:22,853 those restores, depending on what you're restoring, like I could 308 00:15:22,853 --> 00:15:28,073 see a file system maybe not being a big deal for doing the 60 days. 309 00:15:28,078 --> 00:15:30,853 I'm just wondering like database backups, right. 310 00:15:30,853 --> 00:15:33,563 To do those incremental restores. 311 00:15:35,633 --> 00:15:36,263 W. Curtis Preston: Yeah. 312 00:15:36,533 --> 00:15:39,473 Um, I think that's a val, I think that's a valid. 313 00:15:40,328 --> 00:15:41,678 Uh, concern, right? 314 00:15:41,678 --> 00:15:44,678 This is why to this day we still do more frequent 315 00:15:45,248 --> 00:15:46,358 Prasanna Malaiyandi: Fulls that need it. 316 00:15:46,508 --> 00:15:48,248 W. Curtis Preston: on, on database backups. 317 00:15:48,248 --> 00:15:48,638 Right. 318 00:15:48,788 --> 00:15:50,318 The, that's, that's a good point. 319 00:15:50,528 --> 00:15:54,038 Um, and since a significant portion of the world is on, is, 320 00:15:55,958 --> 00:16:01,178 is structured data then, uh, yeah, that's a really good point, right? 321 00:16:02,138 --> 00:16:05,138 So, um, the. 322 00:16:06,488 --> 00:16:10,178 So now we've got a just some different types of, so we're done with the levels. 323 00:16:10,898 --> 00:16:12,848 Now we're talking about different types of backups. 324 00:16:12,848 --> 00:16:15,158 What about a block level incremental backup? 325 00:16:15,788 --> 00:16:16,508 What is that? 326 00:16:18,638 --> 00:16:20,348 Prasanna Malaiyandi: Block level, incremental. 327 00:16:22,208 --> 00:16:24,578 Are you referring to like database slash v. 328 00:16:26,438 --> 00:16:30,098 Where you're just copying the differences in whatever the application 329 00:16:30,098 --> 00:16:34,568 block size is, rather than block backing up the entire image or 330 00:16:34,568 --> 00:16:36,578 the entire file potentially again. 331 00:16:37,208 --> 00:16:40,448 W. Curtis Preston: Yeah, so basically this is the difference between, and 332 00:16:40,448 --> 00:16:43,748 again, this really only applies if you're looking at things at the block level, 333 00:16:43,748 --> 00:16:47,618 which you tend to do at the database, but there are some file systems that 334 00:16:47,618 --> 00:16:50,558 are so dense that backing it up at the block level would be faster than 335 00:16:50,558 --> 00:16:53,108 backing it up at the, at the file level. 336 00:16:53,348 --> 00:16:53,708 Prasanna Malaiyandi: Yeah. 337 00:16:54,248 --> 00:16:57,008 W. Curtis Preston: And so you can do a block level incremental, basically. 338 00:16:57,008 --> 00:16:59,108 The point is, when I'm doing an incremental. 339 00:16:59,813 --> 00:17:02,663 A file system in increment, a file system incremental. 340 00:17:02,663 --> 00:17:05,753 I'm backing up the files that have changed since the full, if I'm doing 341 00:17:05,753 --> 00:17:09,323 a block level incremental, I'm backing up the blocks that have changed since 342 00:17:09,323 --> 00:17:11,813 the full or since the previous bite. 343 00:17:11,963 --> 00:17:12,293 Right. 344 00:17:13,073 --> 00:17:17,303 And then here's something that hopefully you'll know something about, and it's 345 00:17:17,303 --> 00:17:18,683 called Source side deduplication. 346 00:17:19,343 --> 00:17:21,503 This is another way to do incremental. 347 00:17:22,073 --> 00:17:23,033 What can you tell us about. 348 00:17:23,468 --> 00:17:23,708 Prasanna Malaiyandi: Yeah. 349 00:17:23,738 --> 00:17:25,238 Source side de-duplication. 350 00:17:25,478 --> 00:17:31,388 So when de-duplication appliances came out, um, what you ended up happening 351 00:17:31,388 --> 00:17:34,388 is whenever you would do a full right, you were talking right Curtis, about 352 00:17:34,393 --> 00:17:36,638 doing weekly fulls, daily incrementals. 353 00:17:36,878 --> 00:17:40,568 Every time you did that full, probably 98% of the data was the same. 354 00:17:40,898 --> 00:17:45,158 So on the source you would read it, you would send it over the network, it would 355 00:17:45,158 --> 00:17:48,788 go to your de-duplicated appliance and it would throw away 98% of the data. 356 00:17:49,388 --> 00:17:53,793 And so source side de-duplication is, Instead of sending all of 357 00:17:53,793 --> 00:17:54,963 that data to be thrown away. 358 00:17:56,003 --> 00:18:00,863 On the appliance, you basically split the deduplication algorithm 359 00:18:00,863 --> 00:18:03,473 such that you're running some of the processing on the source. 360 00:18:03,713 --> 00:18:07,613 So you figure out what is duplicate, usually at a smaller block level 361 00:18:07,613 --> 00:18:12,203 or some other, uh, granular size and only send the unique blocks 362 00:18:12,208 --> 00:18:14,123 over to the target appliance. 363 00:18:14,453 --> 00:18:17,153 And the rest of the data is sort of reconstructed, if you will. 364 00:18:17,153 --> 00:18:18,683 There are different methods you could look at. 365 00:18:19,218 --> 00:18:23,178 It's sort of rebuilt because you still need to account for those pieces 366 00:18:23,178 --> 00:18:26,588 of data, even though you haven't technically sent the data over the wire. 367 00:18:27,443 --> 00:18:30,593 W. Curtis Preston: Yeah, so if we think about it, like the only reason 368 00:18:30,593 --> 00:18:36,053 ever to do an incremental backup is to save something, save bandwidth, 369 00:18:36,058 --> 00:18:38,303 save CPU time, save storage. 370 00:18:39,533 --> 00:18:43,613 And what, you know, what you're talking about is even when you had 371 00:18:43,618 --> 00:18:48,473 target site dation with like data domain and the like, you, you. 372 00:18:50,183 --> 00:18:51,413 , you save storage. 373 00:18:51,443 --> 00:18:53,783 You, you can do full backup every day and you save storage. 374 00:18:53,783 --> 00:18:57,563 But you were still using up all that CPU in that network by doing 375 00:18:57,568 --> 00:19:00,683 Source I duplication and making that decision before you send it. 376 00:19:00,683 --> 00:19:05,543 You save CPU time, you save bandwidth, you save storage, you save all the things. 377 00:19:05,633 --> 00:19:05,993 Right? 378 00:19:06,563 --> 00:19:12,173 Um, source ID is not as prevalent as targets. 379 00:19:12,173 --> 00:19:12,613 I dedup. 380 00:19:13,043 --> 00:19:14,963 There's, there's only a handful of companies that, 381 00:19:15,023 --> 00:19:16,973 that do it, but uh, they are, 382 00:19:17,783 --> 00:19:17,963 Prasanna Malaiyandi: Yep. 383 00:19:18,353 --> 00:19:18,923 W. Curtis Preston: Um, 384 00:19:19,193 --> 00:19:23,243 Prasanna Malaiyandi: it does require more integration into the backup product on the 385 00:19:23,243 --> 00:19:25,013 source in order to take advantage of it. 386 00:19:25,043 --> 00:19:26,333 W. Curtis Preston: exactly right. 387 00:19:26,993 --> 00:19:30,773 Uh, for the record, Druva is a source site duplication product. 388 00:19:31,493 --> 00:19:36,353 So the next thing we talk about here is called synthetic full backups, 389 00:19:37,703 --> 00:19:40,103 uh, which is and interesting. 390 00:19:40,103 --> 00:19:43,403 I remember when this concept first came out, again, we're trying to 391 00:19:43,893 --> 00:19:45,383 minimize the impact of the backup. 392 00:19:46,773 --> 00:19:48,128 on the system, we're backing up. 393 00:19:48,128 --> 00:19:54,788 So the idea was we already have all of the bites necessary to create a full backup. 394 00:19:54,788 --> 00:19:59,258 Why do we need to go get, you know, we're doing a, we're doing an increment. 395 00:19:59,263 --> 00:20:03,098 We're doing the next backup and 99% of what's on the system we 396 00:20:03,103 --> 00:20:05,678 already have on disc somewhere. 397 00:20:05,768 --> 00:20:07,748 This is really a disc thing. 398 00:20:08,798 --> 00:20:11,468 Uh, we did it on tape too, but it was a lot more work. 399 00:20:11,793 --> 00:20:11,978 Prasanna Malaiyandi: Yeah. 400 00:20:12,728 --> 00:20:15,638 W. Curtis Preston: So we'll go get the 1% that has changed. 401 00:20:16,723 --> 00:20:20,513 and then we'll create a full backup by copying the stuff we already 402 00:20:20,513 --> 00:20:23,603 have and the stuff that we just got. 403 00:20:23,993 --> 00:20:26,423 So we create a synthetic full. 404 00:20:26,873 --> 00:20:29,873 Um, have you, have you run into those out in the wild? 405 00:20:30,278 --> 00:20:33,488 Prasanna Malaiyandi: no, it's very common, uh, especially when 406 00:20:33,488 --> 00:20:35,618 you think about VMware images. 407 00:20:36,113 --> 00:20:36,723 W. Curtis Preston: Mm-hmm. 408 00:20:36,823 --> 00:20:39,908 Prasanna Malaiyandi: depending on how it's being backed up, a lot of folks 409 00:20:39,908 --> 00:20:44,558 would back up a VM as incrementals and then, because like you said, most of 410 00:20:44,558 --> 00:20:49,208 the data is the same, they would sort of synthesize a full image on the target 411 00:20:49,208 --> 00:20:52,118 side in order to have that full copy. 412 00:20:52,118 --> 00:20:54,578 Therefore, when you need to restore, you have that. 413 00:20:54,578 --> 00:20:58,508 What image you could pull from or would a lot of people start doing is you take 414 00:20:58,508 --> 00:21:01,328 that one image and you can now mount. 415 00:21:02,323 --> 00:21:02,528 , right. 416 00:21:02,618 --> 00:21:04,518 And access it directly. 417 00:21:04,523 --> 00:21:06,638 Instant access is what simple some people call it. 418 00:21:06,908 --> 00:21:09,578 So you can spin up the VM from your target system and then 419 00:21:09,818 --> 00:21:11,618 vMotion it over or whatever else. 420 00:21:11,618 --> 00:21:15,758 So you have shorter RTOs than trying to first restore the 421 00:21:15,758 --> 00:21:16,838 data before you spin it up. 422 00:21:18,458 --> 00:21:20,078 W. Curtis Preston: and there, and there's a couple different ways that 423 00:21:20,078 --> 00:21:21,458 you can create a synthetic full. 424 00:21:21,458 --> 00:21:24,878 You can either do it by copying, which is what I was talking about before. 425 00:21:25,178 --> 00:21:27,878 There's also this concept of virtual synthetic full, 426 00:21:28,238 --> 00:21:30,148 where you can just create one. 427 00:21:30,683 --> 00:21:32,468 , you know, magically, right? 428 00:21:32,468 --> 00:21:35,138 Like the, I know that again, data domain I know supported 429 00:21:35,138 --> 00:21:36,248 that on the back end, right? 430 00:21:36,248 --> 00:21:39,908 That if the backup product could say, you have all the bites you need to make a 431 00:21:39,908 --> 00:21:42,938 full, why don't you just put 'em together and, you know, everything's beautiful, 432 00:21:43,523 --> 00:21:44,723 Prasanna Malaiyandi: You know what else does it too, 433 00:21:45,308 --> 00:21:45,918 W. Curtis Preston: what's that? 434 00:21:46,463 --> 00:21:47,063 Prasanna Malaiyandi: Oracle? 435 00:21:48,743 --> 00:21:49,913 When you do incremental merch, 436 00:21:50,648 --> 00:21:51,128 W. Curtis Preston: That's right. 437 00:21:51,128 --> 00:21:52,448 Incremental merge creates. 438 00:21:52,508 --> 00:21:53,978 Exactly, exactly. 439 00:21:54,698 --> 00:21:56,198 Um, but here's the thing. 440 00:21:58,768 --> 00:22:00,938 Wherever we can get rid of full backups, I think we should get rid 441 00:22:00,938 --> 00:22:06,158 of full backups, Right, because they're just the dumbest thing. 442 00:22:06,188 --> 00:22:06,548 Right? 443 00:22:06,548 --> 00:22:10,538 They're just, they're just, they're, we did 'em because that's the way we did 'em. 444 00:22:10,538 --> 00:22:13,008 And we still do 'em because that's the way we used to do 'em. 445 00:22:13,688 --> 00:22:16,118 Uh, so I'm a fan of, go ahead. 446 00:22:16,778 --> 00:22:19,778 Prasanna Malaiyandi: I, so I, I wanna be careful, right? 447 00:22:20,228 --> 00:22:21,368 I think we should. 448 00:22:22,553 --> 00:22:27,953 Get rid of the concept of backing up full copies of data every 449 00:22:27,953 --> 00:22:29,873 single time or on some increment. 450 00:22:30,263 --> 00:22:32,033 I a hundred percent agree with that statement. 451 00:22:32,453 --> 00:22:33,693 W. Curtis Preston: Isn't that what I just said? 452 00:22:33,893 --> 00:22:34,684 Prasanna Malaiyandi: No, no, no, no, no. 453 00:22:34,689 --> 00:22:39,953 But, but I don't agree with, but there's still a necessity to have the 454 00:22:39,953 --> 00:22:42,563 equivalent of periodic fulls, right? 455 00:22:42,683 --> 00:22:45,713 Something that represents a full backup. 456 00:22:46,313 --> 00:22:50,843 Now it could be stored on de-duplicated systems to save storage space, cuz 457 00:22:50,843 --> 00:22:54,863 most of it's the same, but I'm just saying you don't wanna go the route of, 458 00:22:55,073 --> 00:23:00,113 it's just a full once a year and just incrementals physically stored on disk. 459 00:23:00,653 --> 00:23:05,663 So using virtual synthetics or some other mechanism to end up with full copies 460 00:23:05,663 --> 00:23:07,973 periodically, I think is beneficial. 461 00:23:09,398 --> 00:23:12,128 W. Curtis Preston: So I would argue that that's what Incremental Forever 462 00:23:12,128 --> 00:23:14,288 is a true incremental forever system. 463 00:23:14,648 --> 00:23:15,068 Right. 464 00:23:15,608 --> 00:23:19,178 That basically you, which is what, what I was about to talk 465 00:23:19,203 --> 00:23:19,693 Prasanna Malaiyandi: Okay. 466 00:23:19,693 --> 00:23:20,173 Okay. 467 00:23:21,208 --> 00:23:26,738 W. Curtis Preston: Um, and that, you know, the idea is store each backup 468 00:23:26,918 --> 00:23:34,658 in such a way so that a, any restore from any backup, you just, the, it's 469 00:23:34,663 --> 00:23:39,198 stored in such a way that it looks like a full right and a restore from it 470 00:23:39,218 --> 00:23:44,648 behaves like a full, I, I don't think we need to keep making fulls or make or 471 00:23:45,238 --> 00:23:47,148 synthetically creating an occasional full. 472 00:23:47,168 --> 00:23:52,148 I think that if you design it from the beginning, so that you do 473 00:23:52,148 --> 00:23:56,558 incremental forever, and then you store that data in such a way that 474 00:23:57,068 --> 00:23:59,438 every backup is essentially a full 475 00:23:59,733 --> 00:24:00,153 Prasanna Malaiyandi: Yep. 476 00:24:00,578 --> 00:24:02,708 W. Curtis Preston: from, from a way it behaved the way it behaves 477 00:24:02,708 --> 00:24:07,448 during a restore, then I think that's as good as it's gonna get. 478 00:24:09,968 --> 00:24:10,448 Right. 479 00:24:10,643 --> 00:24:15,713 Prasanna Malaiyandi: I, I think that works in the cases of file 480 00:24:15,713 --> 00:24:17,363 systems and virtual machines. 481 00:24:18,023 --> 00:24:23,573 I do think in the case of databases, it is a little database is always that oddball. 482 00:24:23,683 --> 00:24:26,168 W. Curtis Preston: I had the caveat back earlier where I was 483 00:24:26,168 --> 00:24:29,018 like, wherever we can, right. 484 00:24:29,318 --> 00:24:30,218 Um, 485 00:24:30,593 --> 00:24:33,503 Prasanna Malaiyandi: Databases have so many nuances with fulls 486 00:24:33,503 --> 00:24:37,943 and incrementals and log archived, redo logs, and all the rest that I 487 00:24:37,948 --> 00:24:41,483 think, yeah, for everything else, a hundred percent agree with you, Curtis 488 00:24:42,398 --> 00:24:42,938 W. Curtis Preston: Yeah. 489 00:24:43,028 --> 00:24:47,018 Um, and there are, there are a handful of products that back up this way 490 00:24:47,018 --> 00:24:53,153 that basically, said full backups is so from the tape land, right? 491 00:24:53,153 --> 00:24:54,443 Except for database backups. 492 00:24:54,443 --> 00:24:54,713 Right. 493 00:24:55,343 --> 00:24:58,733 Um, generally speaking, there're gonna be products that have come 494 00:24:58,738 --> 00:25:00,803 out in the last 10 ish years. 495 00:25:01,283 --> 00:25:01,673 Right. 496 00:25:01,853 --> 00:25:05,123 Um, products that have been around for 20 years. 497 00:25:05,543 --> 00:25:08,603 They're gonna have spent time in the tape land and they're gonna have 498 00:25:08,603 --> 00:25:11,153 parts of their architecture that are left over from the tape plan. 499 00:25:11,678 --> 00:25:15,158 Prasanna Malaiyandi: I also do wonder if some of that is just the 500 00:25:15,158 --> 00:25:17,168 people who are banishing the systems. 501 00:25:17,198 --> 00:25:23,918 It's something that they are used to, and so change is hard, you know? 502 00:25:24,293 --> 00:25:24,833 W. Curtis Preston: is hard. 503 00:25:24,838 --> 00:25:26,243 Inertia is something. 504 00:25:26,633 --> 00:25:27,833 Yeah, exactly. 505 00:25:28,943 --> 00:25:33,203 And I would say like if you, if you're doing source side dup with incremental 506 00:25:33,203 --> 00:25:37,493 forever, I think from an efficiency perspective, assuming that the way you 507 00:25:37,493 --> 00:25:43,553 store it, everything behaves like a full, an efficiency perspective, I don't see it. 508 00:25:43,613 --> 00:25:46,193 I don't know how it could get any more efficient in that. 509 00:25:46,198 --> 00:25:46,533 Right. 510 00:25:47,123 --> 00:25:51,693 You could maybe argue the block level incremental forever where you're. 511 00:25:52,508 --> 00:25:52,958 Right. 512 00:25:52,958 --> 00:25:57,368 So there, there are like c d P and near CDP systems that's continuous data 513 00:25:57,368 --> 00:26:02,378 protection where they're doing block level changes and they're replicating that. 514 00:26:02,378 --> 00:26:03,728 That's also quite efficient. 515 00:26:03,728 --> 00:26:06,788 It's not source I, it's still, it's still, so I would say, 516 00:26:07,148 --> 00:26:08,438 so I'll restate my statement. 517 00:26:08,468 --> 00:26:14,018 Either source iddu or block level of replication and then incremental 518 00:26:14,018 --> 00:26:17,378 forever stored in such a way so that everything behaves like a full. 519 00:26:17,663 --> 00:26:18,053 Prasanna Malaiyandi: Yeah. 520 00:26:19,558 --> 00:26:21,698 W. Curtis Preston: you know, that includes a handful of 521 00:26:21,698 --> 00:26:24,078 products , one of which is Druva. 522 00:26:24,098 --> 00:26:28,208 But, um, you know, there, there are other products that behave like that. 523 00:26:28,688 --> 00:26:31,388 Like, you know, your old employer, NetApp, right? 524 00:26:31,388 --> 00:26:33,938 That was a block level incremental replication. 525 00:26:34,688 --> 00:26:35,798 Um, yeah. 526 00:26:36,368 --> 00:26:42,368 Um, so I'm gonna throw out one, um, this is gonna be one of our shorter episodes. 527 00:26:42,368 --> 00:26:42,728 I think. 528 00:26:42,758 --> 00:26:43,438 We'll see. 529 00:26:43,808 --> 00:26:44,468 We're almost done. 530 00:26:45,248 --> 00:26:48,188 I'm gonna throw out one, like, uh, what's the cranky old man? 531 00:26:48,938 --> 00:26:49,598 Comment 532 00:26:49,883 --> 00:26:51,173 Prasanna Malaiyandi: get off my lawn. 533 00:26:52,208 --> 00:26:55,088 W. Curtis Preston: this is the archive bit and Windows. 534 00:26:57,278 --> 00:27:00,548 I hate the archive bit and windows. 535 00:27:01,298 --> 00:27:03,098 Um, so what is the archive bit? 536 00:27:03,578 --> 00:27:07,238 It's a flag on every file that when your backup software backs up 537 00:27:07,238 --> 00:27:08,828 that file, it can unset that flag. 538 00:27:09,098 --> 00:27:12,788 If the file is new or it changes, the archive bit is set. 539 00:27:13,088 --> 00:27:16,718 And then when the flag, uh, when the back, the, uh, backup software backs 540 00:27:16,718 --> 00:27:18,968 it up, the archive flag is unset. 541 00:27:19,838 --> 00:27:20,588 Prasanna Malaiyandi: Why. 542 00:27:22,268 --> 00:27:26,198 W. Curtis Preston: This is the old way and it's still the current way. 543 00:27:26,378 --> 00:27:27,458 Uh, not looking. 544 00:27:27,518 --> 00:27:30,218 This is if you're doing a file system level backup in Windows. 545 00:27:31,658 --> 00:27:36,278 Um, I think it's so dumb for multiple reasons. 546 00:27:37,088 --> 00:27:39,038 The first of which it's name, 547 00:27:40,358 --> 00:27:40,988 Prasanna Malaiyandi: Archive. 548 00:27:41,828 --> 00:27:43,058 W. Curtis Preston: it should be the backup bit. 549 00:27:43,478 --> 00:27:43,718 Prasanna Malaiyandi: Yeah. 550 00:27:44,168 --> 00:27:47,348 W. Curtis Preston: They called it the archive bit anyway, but that's not 551 00:27:47,348 --> 00:27:55,733 really my problem is you can't, if anyone comes along with like, any, cuz there 552 00:27:55,733 --> 00:28:00,833 are a number of like third party tools that can back up your Windows system. 553 00:28:02,273 --> 00:28:06,233 If you run a third party tool to create your own backup, like let's 554 00:28:06,233 --> 00:28:10,163 say you, you know, a lot, some people don't trust the backup system, right? 555 00:28:10,913 --> 00:28:16,913 So if you, if you being an admin go and decide to make your own backup 556 00:28:17,063 --> 00:28:21,053 and you use something that uses the archive bit, it will clear the 557 00:28:21,053 --> 00:28:23,753 archive bit on all the files that have changed since the last full. 558 00:28:25,038 --> 00:28:29,153 and then when you go, when the real backup software comes in, it actually 559 00:28:29,333 --> 00:28:29,903 Prasanna Malaiyandi: not gonna find, 560 00:28:30,083 --> 00:28:31,313 W. Curtis Preston: until the next full backup. 561 00:28:32,603 --> 00:28:33,953 Prasanna Malaiyandi: and I bet people don't 562 00:28:33,983 --> 00:28:34,673 W. Curtis Preston: problem with it. 563 00:28:34,853 --> 00:28:35,213 Prasanna Malaiyandi: Yeah. 564 00:28:35,693 --> 00:28:35,963 W. Curtis Preston: Yeah. 565 00:28:35,963 --> 00:28:37,223 And of course people don't notice it. 566 00:28:37,613 --> 00:28:46,463 Um, and so I'm gonna go back to the story, um, that I talked about. 567 00:28:46,673 --> 00:28:50,693 I don't know what the, the one about me being kidnapped, was 568 00:28:50,693 --> 00:28:51,623 that in this episode or was 569 00:28:51,628 --> 00:28:51,773 That 570 00:28:51,788 --> 00:28:52,508 Prasanna Malaiyandi: was a previous. 571 00:28:53,598 --> 00:28:59,573 W. Curtis Preston: Okay, well I hinted at this story from um, this other, 572 00:28:59,573 --> 00:29:04,703 from the other episode, but here's the full story cuz we got time And you know 573 00:29:04,703 --> 00:29:07,973 what, if you don't want another old backup story then you know, thanks for 574 00:29:07,973 --> 00:29:10,373 joining and uh, see you next episode. 575 00:29:10,703 --> 00:29:17,333 But this is a good story and if figures in a good buddy of mine, Rob Worman. 576 00:29:17,813 --> 00:29:20,153 I know Rob listens to the podcast. 577 00:29:20,783 --> 00:29:22,193 , he figures into this story. 578 00:29:22,763 --> 00:29:29,813 So I was at this large, you know, national entertainment company and I was helping 579 00:29:29,818 --> 00:29:31,313 them to redesign their backup system. 580 00:29:31,318 --> 00:29:33,473 We were doing a massive redesign. 581 00:29:34,523 --> 00:29:41,303 And one of the, and one of the things, you know, like we pushed them from, you 582 00:29:41,303 --> 00:29:42,983 know, weekly folds to monthly folds. 583 00:29:42,983 --> 00:29:44,963 We, we changed all their multiplexing setting. 584 00:29:44,963 --> 00:29:48,023 We changed what, uh, we changed how. 585 00:29:48,773 --> 00:29:52,133 , the schedules were, I mean, we changed everything and we were doing this 586 00:29:52,133 --> 00:29:54,713 to, to do massive efficiency change. 587 00:29:54,713 --> 00:30:02,033 They were running like, it was like 18 tape drives simultaneously, 588 00:30:02,063 --> 00:30:05,603 and I told 'em that that was their pro, they were gonna buy, uh, two 589 00:30:05,738 --> 00:30:08,018 Prasanna Malaiyandi: Oh, this is a sh when you're, you were, you're talking about the 590 00:30:08,018 --> 00:30:12,488 shoe shining problem and they were trying to tape goes too fast for what they need. 591 00:30:13,073 --> 00:30:13,553 W. Curtis Preston: Yeah. 592 00:30:13,583 --> 00:30:17,783 And so they were gonna buy two new tape drives and that was gonna fix the problem. 593 00:30:17,783 --> 00:30:18,803 And I'm like, no, it's not. 594 00:30:18,803 --> 00:30:19,823 It's actually gonna make it worse. 595 00:30:19,823 --> 00:30:23,423 And I explained to 'em and I'm like, listen, give me a few weeks and I'll 596 00:30:23,773 --> 00:30:24,683 redesign your backups to submit. 597 00:30:24,688 --> 00:30:27,083 It'll cost you less than what those tape drives were gonna cost. 598 00:30:27,083 --> 00:30:28,133 Cuz they were expensive. 599 00:30:28,523 --> 00:30:32,363 These were the storage tech tape drives, like the 90, what? 600 00:30:32,363 --> 00:30:33,743 I can't remember the tape name. 601 00:30:33,803 --> 00:30:34,283 Tape names. 602 00:30:34,283 --> 00:30:40,373 But anyway, um, so we did this and. 603 00:30:41,273 --> 00:30:52,238 , um, and things went really well, but so I knew about the archive bit. 604 00:30:52,268 --> 00:30:53,528 This is all about the archive bit. 605 00:30:53,558 --> 00:31:00,338 I knew about the archive bit and I knew that if I had, I basically 606 00:31:00,338 --> 00:31:02,858 left their parallel backup system. 607 00:31:03,038 --> 00:31:04,388 I left their parallel policy. 608 00:31:06,023 --> 00:31:06,323 Prasanna Malaiyandi: Up and 609 00:31:06,623 --> 00:31:09,863 W. Curtis Preston: I left their existing policies running in parallel. 610 00:31:10,103 --> 00:31:13,703 Well, I was gonna do the regular backups, uh, during the day, 611 00:31:14,843 --> 00:31:17,933 um, and something like that. 612 00:31:17,993 --> 00:31:18,623 It's been a while. 613 00:31:19,943 --> 00:31:25,553 And, but I knew that I would screw them up if I had the archive bid on both systems. 614 00:31:25,553 --> 00:31:27,623 So I used, this was net backup. 615 00:31:27,833 --> 00:31:34,403 I used a feature of net backup that said, okay, don't use the archive bit on. 616 00:31:35,198 --> 00:31:41,978 On the sink and what I didn't notice or what happened over time, over about 617 00:31:41,978 --> 00:31:45,458 two weeks is the arc, the incremental backups got bigger and bigger and 618 00:31:45,458 --> 00:31:47,918 bigger and bigger, and they got huge 619 00:31:48,288 --> 00:31:48,563 Prasanna Malaiyandi: on your new 620 00:31:48,848 --> 00:31:51,998 W. Curtis Preston: and they on the new system and they got so 621 00:31:51,998 --> 00:31:53,468 big that they weren't finishing. 622 00:31:53,558 --> 00:31:56,648 By the way, when I got there, they were backing up 28 hours a day. 623 00:31:57,038 --> 00:32:01,718 So my, my opinion was no matter what I was doing, it still wasn't gonna be that bad. 624 00:32:04,333 --> 00:32:08,558 , but backup, but the backups weren't finishing and I didn't understand, right. 625 00:32:08,558 --> 00:32:10,298 So, so we shut off the old backups. 626 00:32:10,298 --> 00:32:13,328 We're like, okay, we we're, we're beyond the point of no return. 627 00:32:13,328 --> 00:32:15,968 We shut off the old backups, backups still weren't finishing. 628 00:32:16,868 --> 00:32:20,258 And uh, you know, I was working through diff, you know, 629 00:32:20,258 --> 00:32:21,668 multiple levels of support. 630 00:32:21,668 --> 00:32:24,488 And I actually at, at one point ended up with my buddy Rob, 631 00:32:24,968 --> 00:32:25,478 Prasanna Malaiyandi: Hmm. 632 00:32:26,948 --> 00:32:28,148 W. Curtis Preston: I'm like, dude, I don't understand. 633 00:32:28,448 --> 00:32:31,748 And we, you know, we figured out that when that. 634 00:32:32,663 --> 00:32:34,343 Doesn't use the archive bit. 635 00:32:34,373 --> 00:32:38,333 What it does is a, again, this is, this is a hundred years ago. 636 00:32:38,333 --> 00:32:39,263 I don't know what it does today. 637 00:32:40,973 --> 00:32:44,603 If it doesn't use the archive bit, it would, it would traverse the directory 638 00:32:44,603 --> 00:32:51,713 tree and then it would, as soon as it encountered a directory that whose 639 00:32:51,713 --> 00:32:55,463 modification time had changed, it would back up everything in that directory tree, 640 00:32:56,868 --> 00:32:57,173 Prasanna Malaiyandi: Okay. 641 00:32:58,313 --> 00:33:00,533 Including data that does, yeah, 642 00:33:01,003 --> 00:33:03,578 W. Curtis Preston: including data that hadn't changed, right? 643 00:33:03,578 --> 00:33:05,168 It, it was really inefficient. 644 00:33:05,198 --> 00:33:09,548 And so it tried to add that level of efficiency by working up the, 645 00:33:09,878 --> 00:33:12,878 you know, so long story short, the back, the incremental backups were, 646 00:33:12,998 --> 00:33:14,468 were getting close to full backups. 647 00:33:14,473 --> 00:33:17,438 And so once we figured that out, we're like, okay, okay, 648 00:33:17,498 --> 00:33:19,538 okay, okay, we figured it out. 649 00:33:19,748 --> 00:33:21,548 We're gonna tell it to use the archive bit now. 650 00:33:23,108 --> 00:33:27,488 And, um, and we turned on the, we turned on the archive bit. 651 00:33:28,648 --> 00:33:30,098 and it was the same the next day. 652 00:33:30,758 --> 00:33:31,088 Right? 653 00:33:31,493 --> 00:33:31,943 Prasanna Malaiyandi: Okay, 654 00:33:32,258 --> 00:33:34,778 W. Curtis Preston: Now I look, now I look, and meanwhile this, this 655 00:33:34,778 --> 00:33:38,648 manager, he is looking at me, you know, looking at me over the, you know, he's 656 00:33:38,653 --> 00:33:40,478 beginning to think I'm a complete moron. 657 00:33:41,168 --> 00:33:48,188 And, um, he, um, I said, look, I, you know, we figured it out. 658 00:33:48,188 --> 00:33:51,938 This is what, you know, he's like, but it, but it didn't fix the problem. 659 00:33:52,148 --> 00:33:53,708 And that's when I was really talking to Rob. 660 00:33:53,708 --> 00:33:54,398 I'm like, I don't understand. 661 00:33:54,398 --> 00:33:56,678 And again, I was on the edge here. 662 00:33:56,678 --> 00:34:01,688 I would, no one had done what I was doing, you know, and so, so what we 663 00:34:01,688 --> 00:34:06,248 figured out was when you tell net backup not to use the archive bit, 664 00:34:06,938 --> 00:34:08,498 it doesn't use the archive bit. 665 00:34:08,503 --> 00:34:11,528 Now, what that means is that when it backs up a file, it doesn't 666 00:34:11,533 --> 00:34:15,158 clear the archive bit, which was the whole point of me using the thing. 667 00:34:15,788 --> 00:34:17,078 And so that meant that 668 00:34:18,278 --> 00:34:18,398 Prasanna Malaiyandi: The 669 00:34:18,608 --> 00:34:19,388 W. Curtis Preston: we, 670 00:34:19,778 --> 00:34:20,798 Prasanna Malaiyandi: when you turn it on 671 00:34:20,948 --> 00:34:23,498 W. Curtis Preston: to run the, we had to run the one backup. 672 00:34:24,683 --> 00:34:27,623 that would use the archive bit to clear everything. 673 00:34:27,683 --> 00:34:31,643 But because we had run one in a long time, it was also gonna be really long. 674 00:34:31,973 --> 00:34:36,443 And so then it's like, okay, tomorrow night is gonna be all better. 675 00:34:37,433 --> 00:34:45,083 And that's when the boss said, okay, smarty pants, no one's going anywhere. 676 00:34:45,443 --> 00:34:48,773 He was, he was really upset at the level of instability. 677 00:34:48,773 --> 00:34:49,823 And I can understand him being. 678 00:34:50,543 --> 00:34:54,263 But he was really upset at the level of instability in his mind that 679 00:34:54,263 --> 00:34:55,823 I had introduced into his system. 680 00:34:57,143 --> 00:35:01,493 Nevermind that the new backup system was backing up 50% more 681 00:35:01,493 --> 00:35:07,313 data than the old system because I changed the, the inclusion factor. 682 00:35:07,373 --> 00:35:10,433 And they had, they had, they had left out a lot of really important data. 683 00:35:10,433 --> 00:35:14,753 So it went from 20 terabytes to 30 terabytes and. 684 00:35:15,308 --> 00:35:18,608 Nevermind the fact that we were now creating two copies, whereas before 685 00:35:18,608 --> 00:35:20,258 he wasn't even getting one copy done. 686 00:35:20,708 --> 00:35:22,238 So I was feeling pretty good. 687 00:35:22,328 --> 00:35:23,498 But all he sees is his 688 00:35:23,568 --> 00:35:24,368 Prasanna Malaiyandi: Stability. 689 00:35:24,388 --> 00:35:24,808 Yep. 690 00:35:25,718 --> 00:35:26,198 W. Curtis Preston: Yeah. 691 00:35:26,558 --> 00:35:30,218 And so he's like, well, nobody's going home until, and literally 692 00:35:30,308 --> 00:35:32,408 he just, he stayed with us. 693 00:35:32,413 --> 00:35:33,458 I mean, to his credit, 694 00:35:33,658 --> 00:35:34,148 Prasanna Malaiyandi: Okay. 695 00:35:34,388 --> 00:35:38,768 W. Curtis Preston: stayed with us and like he ordered pizza and stuff, uh, 696 00:35:38,768 --> 00:35:40,688 but he literally wouldn't let us leave. 697 00:35:41,963 --> 00:35:44,003 I couldn't even say like, well, I kicked off the back house. 698 00:35:44,003 --> 00:35:44,783 We're gonna go get dinner. 699 00:35:45,053 --> 00:35:46,223 I ain't nobody going anywhere. 700 00:35:47,603 --> 00:35:48,323 Prasanna Malaiyandi: He's like bathroom. 701 00:35:48,323 --> 00:35:49,433 You already had your bathroom break. 702 00:35:50,483 --> 00:35:52,913 W. Curtis Preston: but we're gonna sit here and Yeah. 703 00:35:52,913 --> 00:35:55,973 And so we were there until pretty late. 704 00:35:56,633 --> 00:36:00,923 Um, I don't really remember how late, but I remember like, it was like a whole 705 00:36:00,923 --> 00:36:06,023 second day, you know, like it was, it was a long, long, long, long night. 706 00:36:06,083 --> 00:36:10,433 And that my friends, . That's what happens when you don't understand backup levels. 707 00:36:11,003 --> 00:36:15,173 Anyway, All right. 708 00:36:15,173 --> 00:36:19,943 Well I think this is enough torturing people with backup level information. 709 00:36:21,683 --> 00:36:23,303 Prasanna Malaiyandi: I, I a hundred percent 710 00:36:23,393 --> 00:36:24,983 W. Curtis Preston: talk about, go ahead. 711 00:36:24,988 --> 00:36:25,343 Yeah. 712 00:36:25,613 --> 00:36:27,413 Next week we're gonna talk about metrics. 713 00:36:27,863 --> 00:36:28,283 Prasanna Malaiyandi: Ooh, 714 00:36:29,063 --> 00:36:29,423 W. Curtis Preston: yeah. 715 00:36:29,873 --> 00:36:31,133 I think it's a little more exciting. 716 00:36:31,373 --> 00:36:31,673 Prasanna Malaiyandi: yep. 717 00:36:32,603 --> 00:36:33,323 W. Curtis Preston: It's backups. 718 00:36:33,323 --> 00:36:34,723 Like how exciting could it be 719 00:36:35,798 --> 00:36:36,998 Prasanna Malaiyandi: Come on, Curtis. 720 00:36:36,998 --> 00:36:38,018 What are you talking? 721 00:36:38,168 --> 00:36:40,598 Telling our audience, our listeners, 722 00:36:41,573 --> 00:36:42,263 W. Curtis Preston: Woo. 723 00:36:42,263 --> 00:36:43,223 I dunno what to tell you. 724 00:36:43,463 --> 00:36:46,163 You know what, maybe you're excited as you're as excited 725 00:36:46,163 --> 00:36:50,933 about backups as I am, maybe, to which I say, welcome to the party. 726 00:36:51,803 --> 00:36:55,403 All right, well, uh, thanks for listening folks, and remember to 727 00:36:55,403 --> 00:36:57,623 subscribe so that you can restore it all.