Skip to Content
59:09 Webinar

Your Personal Database Time Machine: The SQL

This month, host Andrew Miller will invite SQL DBA and Microsoft MVP Anthony Nocentino to the virtual break room to discuss all things SQL Server landscape, SQL Server 2022, and Pure Storage.
This webinar first aired on Sept. 6, 2022 | 11:00am PT | 2:00pm ET | 60 mins
The first 5 minute(s) of our recorded Webinars are open; however, if you are enjoying them, we’ll ask for a little information to finish watching.
Click to View Transcript
00:00
Thank you everybody for joining us for this month, September of 2022 coffee break. I'm your host, Andrew Miller, joined by Anthony Noc Andino. I think by now I've managed to say your name right? Or maybe wrong a lot of times. But you know, Anthony, thank you so much for joining us today. We're gonna be covering your personal database
00:16
time machine, the sequel or SQL S we were debating beforehand. This is gonna be both about pure and about all things sequel and even a little bit of Anthony's history in coming to pure as always. This is a series we are now on month. Oh, I think it's 21 or so. Give or take if you ever want to hear any of the previous months. It's crazy.
00:35
Um And actually, um hearing of the previous months, we actually keep them all online. There's a little bit of an easier URL to remember that bit dot Lee slash Pure Coffee Break recordings. It's not just meant to be like, oh, you have to go listen to the pure stuff, but there's enough solution focus and educational value.
00:52
Actually, I think they've aged pretty well and I've had that feedback from folks. I actually saw some of you potentially give a shout out to the chat if saw you at VM Ware explore this last week. I actually had a couple of folks come up and say hello and want to take pictures and hey, that feels good. You know, it's like it's fun as well. This is,
01:07
there is a little bit of incentive here. After all, there's coffee cards and there will be, we will have a drawing for an ember mug, $130 retail value, the kind you can control your phone at the end. For those of you that are in this these categories, we appreciate you joining. But as I think, you know, we can't send you cards almost finally from a housekeeping
01:26
standpoint. Next month, we will be joined by Josh Gluck who is the Vice President of Global Vertical Alliances and Solutions. That's a mouthful. But what's more important is that he comes from a customer background as a health care, it executive and also finance background. So, you know the how sick is your data platform is playing off that a little bit.
01:43
We're gonna go into Josh's experience as an it customer executive and then how pure helps health care organizations and finance. But actually even more, we're gonna tell some stories or oriented around those areas, but then generalize it a little bit. As always, I'm your host, Andrew Miller, not gonna introduce myself.
02:00
Except to say that when I was on the customer side, I did support DB A s who ran sequel server and even DB two. And for, I don't know if that's gonna make you go like this at me, Anthony. I, I don't know or not. So, um but then mostly did um my sequel and post that was actually out of standing up lamp stuff because we were like, yeah, I wanted to run a forums in Terrible University and some
02:19
other fun stuff. But the reason if we're gonna have a conversation about sequel server and Microsoft, it's far better to have this gentleman cover it than me. Anthony. Do you mind doing a brief introduction? Uh Sure. So, hey, I'm Anthony Nosso, Principal of Field solution architecture here.
02:36
I specialize in system architecture and performance. I like to make things go fast. Uh That's kind of a core part of my job. Uh There's my contact information, I blog a whole bunch at nono dot com. I have a repo where pretty much anything I do publicly lands. So any presentation I've done any code, I've written lands there and I'm also a plural site
02:53
author with a specialization in cnet's and Cloud. Uh Linux and powershell. Uh I've written a bunch of books, MVP and a co organizer of the Fantastic A TV event that um both Andrew Prosky and Mark Wilkinson and I co-founded it a couple of years ago. Want to make sure you gave a shout out for that conference.
03:10
It's very cool at times when you and I have been talking and even doing some preparation for this, we try to keep it relaxed but we prepare, it's respect in the audience. Uh There's been a time where you bring up a top and you're like, oh, what about, it's in this book that I wrote or in this thing that I wrote, you know, or then we, we share o'reilly stories. And what's your favorite o'reilly Animal?
03:24
I don't know, you know, that kind of thing before we dive in though, we got a couple hiding out right there. O'reilly books. Cool. You, you've got the, you've got that better than I do usually like, like the pointing and in reverse. So for those who are joining us, this is the topic.
03:39
Obviously, if you're thinking this feels like a little bit of Deja vu from last year, you're right because about a year ago, actually, it was shortly after you started Anthony and Anthony jumped right in with both feet, was helping with Q and A when our Jeni Fernandez and I went through, oops, I did it again on my SQL server. Yes, that's a little bit of a pop culture reference, you know,
03:57
data protection recovery challenges because of the solution, focus a good bit of that content is still relevant. So I actually encourage you, you know, if you enjoy what we do today, go back and listen to that one. And maybe, maybe if you want to kind of think of it like a two header,
04:10
you know, some of it was around snapshots, there was availability groups. There is actually, by the way, this is a fun one, big data clusters, maybe poor one out for big data clusters. I think Anthony because that was a uh but, but it, it morphed into and help me out there, it morphed into something better when you're talking about. You got the hanky right there too.
04:28
So it, it morphed into. Oh yeah. So I guess it's my Azure enable data services which I actually uh Ben Weisman and I uh wrote two books on um being able to deploy managed has offerings of uh Azure data services like Post Grass and um sequel manage instance uh in hybrid Cloud and on prem scenarios.
04:49
Pretty cool stuff there. So, very neat. So it was a project that it keeps moving. I mean, there's, there's code and all this good stuff. So, and we're right on time as well. I, I've, I've, I don't force the agenda but, you know, folks seem to keep going back. So I usually do a kind of a,
05:03
a simple four part agenda with cheap jokes at the top DB A s are people too. I've met Anthony in person. So I'll vouch that you're a real person. And then so is Melody and other folks, Andy. You. So first we're gonna start off with a little bit of just kind of actually history and even of as we often do here, a little bit of, you know, kind of career,
05:19
maybe, advice and the tipping points and interesting things you've seen. So accidental DB A for life, Anthony's journey, some discussion around SQL server and what's been happening in that landscape overall. So more of an industry focus then we'll dive into. I mean, I don't know about you. But when I say, when I say sequel, you always say three,
05:38
maybe, maybe not. So there's something new here that's fun, actually, very fun to talk about. And then last but not least bending space and time. It's not just snapshots, there's other stuff there as well and we'll do some poll questions along the way, both as kind of palette cleansers and having fun.
05:53
Something that you might even be interested in the results because we'll make sure share the results. They're always anonymous as well. So to dive in. Oh and as well. Um we said it before but Q and A throughout, please make sure to put if you have questions uh like formal Q and A questions,
06:07
put them in the Q and A, feel free in the Zoom chat. We'll try and keep an eye on it. It's usually just a river that keeps going by. And if not, we don't get to your questions during, we'll be done in about 45 minutes after the hour roughly and then we'll just kind of hang out and have a little bit more of a relaxed Q and A, the formal content is done by then.
06:22
And of course, the drawing too if you're hanging out for that. So diving into section number one. So this uh when you said accidental DB A Anthony when we were chatting earlier and you said it feels like this is a, as you express, it is almost a little bit common that people don't like, grow up. And when they're five years old,
06:40
you ask a kid and say, what do you want to do when I grow up? And the answer is something other than DB a right kind of thing. So you might actually kind of walking through a little bit, I think for you, it even kind of started with your college career. I'll just turn it over to you to take it away. So that's for sure.
06:56
Although I'll keep it aside when I took that test in high school, the one that's like, what are you gonna do when you grow up? And it came out systems analyst? So I'm like, you're not too far off, you're not too far true for the broad definition of terms. Yeah. All right.
07:09
So let's see, uh where it starts about college and you were in school? Yeah. Yeah. Yeah. So I went back to college after some college mishaps as a, as a young. Um And I really liked it. So I finished my undergrad. I finished a master's and I started working on a phd um,
07:26
in computer science. And whilst I was working on that, that phd, I had the luxury, this is actually one of the cooler jobs I have. Probably the second coolest job because the coolest job I have is the job I have right now. Uh, the second coolest job I've ever had was I was a network administrator in a computer science department. Um So I was literally the computer nerds,
07:41
computer nerd, right? That was, that was really great. Um And I was a phd student at the same time. Um But around that time, a friend of mine had a developer that quit and he calls me up and he's like, hey, I need somebody to write code to do this thing. I'm like, all right, I'll write code to do that thing.
07:54
Um And I, since I was in college, I, I got that phone call, I was at a bar actually and I called him back and I was like, hey, why don't you just hire me? It's like be your developer and I'll just write all of your code. And he's like, yeah, that's cool. Um So we've worked on this pretty good relationship around that.
08:07
But um and so I got to get in there and I was, I was writing uh electronic medical record software. So tracking specific things around cardiology data was my core disc and epic. Maybe not that big, but in that category at least, maybe. Yeah. Yeah. Yeah. Um And so what I found though is as I wrote code and I stuck data in a database that it
08:26
didn't do things the way that I wanted it to be done fast enough, you know, quick enough because this was, this was health care. So if I match, if a doctor matches a button, I want them to get the answer as fast as possible for obvious reasons. Right? And so, and then I started looking at this, this thing where I was sticking data,
08:42
which was the database uh in this case of a sequel server. And I was like, well, I can do this and make it better and I can do that and make it better. And then what I found out is that inside the database engine itself, it kind of had all the things that I liked, right? I had to worry about like systems and how bits and bytes flew around inside of a computer,
08:57
right? I had to learn how to program or ask the computer to do things, right? And so it was, it was kind of bringing together all these things that I was learning in school, um even in advanced uh graduate school stuff that I could apply directly in real life into what's inside of a relational database engine.
09:12
And I, and that just kind of came my bread and butter after that I kept, kept going deeper and deeper and deeper into, um, how database engines work. And then eventually, um, that company that I worked for got bought by a bigger company and they were like, hey, Anthony, like, yeah, they're like, you're fired. I'm like, what, like, we don't have remote
09:30
employees and I'm like, what do you mean by that? And they're like, well, here's the thing. We'll hire you as a consultant. And I was like, oh, sounds fantastic. Um And so I spent the next couple of years building the business that became Centeno systems. Um Yeah, and it was all downhill since there or
09:45
uphill whatever way is positive. You know, there's even one other earlier origin story and maybe this is just me making cheap jokes from a historical standpoint about if, if anybody remembers back in the days of paper M CS es, you know, everybody and their brother, daughter, dog, cat, et cetera. So that's no, it's all right.
10:01
Everyone had an M CS E kind of thing. But, but even back then, so, I mean, you came into some of this with a real coding background, like you've been researching parallel IO on to and even did some stuff about like writing software similar to a VM ware lockstep. If anybody remembers that. Do you mind just giving a sound bitter or two there?
10:19
Because it's actually kind of cool to hear that you've been doing you know, serious development work even before this. So, yeah. So in graduate school, those were my, my core research projects. Um The initial project was something that was very similar to VM Ware lockstep where I was able to take a set of memories allocated inside the kernel and synchronize that between two
10:38
different independent workstations. And, and the value of that is uh how uh live migration works or uh the motion in the marketing terms uh is it's this iterative process of copying what's called dirty memory pages to host host, right? And I wanted to get rid of the part and make it faster, right? So that you could have immediate fail over and
10:57
also not lose process state in that transition, which is, is the biggest part there. Um And then, yeah, the, the, the one was great in that um I was doing lots of things like IO coalescing distributed IO based on how I thought it was gonna be read later uh with kind of the predecessor to like machine learning. Basically, I could,
11:16
I could kind of classify some patterns that I thought were going to occur in the data sets and then go to retrieve the data in that way. And so those are some of the core things that I worked on uh throughout graduate school. It's actually so last week was at VM Ware Explorer and myself and David Stan were presenting for Tech Field Day and one of the themes that we were bringing out was a lot of
11:35
the work that Pier has been doing around V vas man. We're just outside a sequel now. But oh, well, it's short around batching coalescing and parallel. So it's like you come across these common themes sometimes disaggregated, like these themes that, that kind of wind through all kinds of different technology stuff
11:49
and it plays out here and it plays out here. And so anyway, just made me think about it, especially the parallel. Yeah, in the end, when you talk about getting data between A and B, the idea is if I can get more data in the, the movement from A to B, then the chances are I get more things done, right?
12:04
And that's that con the core concept behind IO coalescing, right? So pulling forward, thanks for letting me kind of go back in time a little, but we jump forward to the company you're with, got bought nomo employees and you've got essentially almost kind of a couple year runway to build a consulting, you know, consulting organization. You might kind of walking through that a little
12:25
bit because there may be some folks on the listening in that have thought about going independent in various ways. And then maybe at whatever point it makes sense to kind of segue into how you got involved in Kuti. So back to you. OK. OK. So let's see. I one of the things I had the luxury of having
12:40
had happen to me is that particular, this is like how to build a consulting business is that I had a two year contract, which I knew that that's when I had a two year contract. But I also had 100% of my consulting revenue coming from one client which is like on the top of things you don't want to have happen when you're running the consulting business, not diversified.
12:59
Yeah, so I spent, I spent the next two years of that contract trying to figure out how to build up other revenue streams from other consulting customers. Uh but also diversifying revenue streams across different ways to generate revenue in that. If consulting goes away, what else was there? And that's where training came in, right? And content publishing.
13:19
Um And so kind of working across both of those angles. I had some stability on uh revenue as a, as a business because one of the things that I found um that I never wanted to do was manage people or also have employees because I felt like I would be personally or financially responsible for other people's families, which is, I don't know, as a burden I could bear as an individual,
13:41
right? And so I just, I was like, you know what, I am going to be the centennial of systems and focus on that. Um I had, I'll tell you the one fun thing that happened during that time. Is, um, I did get to meet Argen. Uh we had a mutual customer around 2015 that we put, I put my first sequel server workload on pure.
13:59
And I was like, wow, that, that's kind of special because all of the numbers that I wanted to be big were big, like throughput and all the numbers that I wanted to be little like latency where they were like, less than zero. Right. Well, less than zero and milliseconds and I had to like, switch the microseconds and I was like, oh, that's nerdy.
14:13
And so it took me a while to get microseconds in my head. But I got it eventually. So. Yeah. Yeah. So that's when I got involved with pure and I was like, wow, this is pretty cool. Um, and ever since then I was a fan, I, I've put a lot of my customers up here when I was doing consulting and honestly, I, I got to become friends with our dentist and many other folks here like melody that,
14:33
uh, I knew from the sequel Surfer Community that are working in pure and I'm like, this just seems like the natural next step, you know, and that's the one I said earlier. I now have the best job I've ever had. And then I could make the joke about how my previous boss was. Yeah, I think that the last thing I want to make sure to pull,
14:51
pull here was a little bit of your involvement in Kubernetes. We already ironically talked about big data clusters and all like imperative, declared it. By the way, for those who joined us last month, we're totally gonna recycle a joke here because everything is solved by saying Cotti or insert whatever buzzword but, but it was a little bit more than buzzword for you as you started to
15:09
get involved there. So, you know, just, just help me out with how DB A and sequel wanders in and links into KTIS. Sure. So I was at a Microsoft event probably 2016, 2017. And Microsoft was talking about big data customs, right?
15:25
And I'm like uh big data classes was cool. But I'm like this, this, this thing, this is, this is gonna be a thing and this is 16 17 because I'm like, I could take code and tell the computer what I wanted to do and it's gonna do it right? Versus the imperative way of doing things where I'm like, I gotta click a button, I gotta write a script. I gotta change the state of a system between A
15:48
and B and then cross my fingers that it stays in that state, right? You know anyone that's done like traditional data center operations, that that's what you've done. You rolled a system out, you've marked it golden, you put it in production, you didn't touch it, right? That's the shift to that.
16:01
That's the shift to declarative techniques where I just write code that describes what I want. I have done and I thought that was such a powerful construct to how we could deploy and manage systems in a data center or cloud. I mean, we get to the idea of item potency. That's your, I guess our $10 word for the day, something like that kind of thing.
16:17
But, but it means something actually about how you update. Um and we're not going by the way, go into it. If you want to cover some of this, go back to maybe last month or even a year ago about state full versus stateless for Kubernetes. I I know you and I talked about, you know, some like, yeah, why can't I just use this tool the way I want to,
16:32
why not bring the compute to the data instead of shoving the data in different places and thinking about stateless? So yeah, that's, oh you want to double click on that? You wanna stay? Yeah, that that was the other part that I felt was super interesting is data lives where it lives.
16:48
If I could just bring a process to where the data is to get access to the data in a container, then that's all I really care about. And then I can focus on what's important is the actual data and decoupling the configuration of the platform from the state of the things that I actually care about is is again a huge powerful concept. I knew this was gonna happen. There's enough cool stuff to talk about.
17:08
We keep going a little longer in the section but we still got time. So before we jump into section number two, Emily, if you don't mind launching poll number one here, um we won't read this too much. So just, just a little bit of, you know, hearing some of Anthony's background. Um You'll see that this is three questions. First, I describe myself as,
17:25
you know, www whatever you identify as right now kind of thing, storage database, et cetera. Uh First choice of SQL server. Um First version of SQL server, we have to combine a couple options together there just because you can only do 10 answers on Zoom. Hopefully that won't require too much mental over from you. And then actually curious to know if people
17:42
have dedicated DB A S, we're just gonna leave this open as we go into section number two. And as we get to the end of section number two, we'll close the pole up and uh and share the answers with you, but moving into section two. So that's a little bit um kind of about you. This is where our power point decides to be a little bit fun. There we go. But if, if we now flip over because a lot of
18:05
your, a lot of your life, although we didn't call out, you know, all the years of consulting that you did was all about sequel server. Um If you had to think about, not that we thought about this ahead of time at all. But if you had to think about a couple of big sequel server trends, I, I think there were three of them that you had in mind.
18:22
Uh the first, which feels almost a little bus worthy, but I think there's reality to it as well, kind of hybrid cloud and making it real. But from a sequel server standpoint, you mind expanding that a little bit. Sure. Sure. So if you look at the 22 major steps that Microsoft has done with SQL server in the last year or two, we have what's called Azure Arable Data services,
18:45
which gives me the ability to run Paz services like SQL M I or even post sequel wherever I have Rene that could be in Azure, that could be in Aws that could be on a shipping container in the middle of the ocean, which is a practical scenario that there um is being serviced, right? And that's a huge thing, right? Because then I could just consume data where I need to consume data.
19:08
The other thing that I see the other definition of hybrid. So that's kind of the one of the core thing. The other big play that you see for Microsoft is specifically in sequel server 22 is where they're, they're making the the transition or they're making the the lines between cloud and on prem blurrier, right? In that I can consume um or produce a data
19:31
either on primer and cloud and have that function interchangeably easily, right? So things like we'll talk about S3 in a minute or things like Azure um Slmim I link where I can get automatic fail over or particularly not automatic failure, but just fail over into Azure with a couple clicks of a mouse inside of SQL server Management studio.
19:49
And that's really been, I think the core focus of Microsoft is is reducing the friction in those two areas to get those kind of hybrid scenarios wherever you really need your data, right? I mean, it feels like it's I'm being a little trite but the whole data in database and it's about data portability and focusing on data gravity and how that creates friction and or impedes getting value out of the data.
20:14
I feel like I'm just repeating back what you said with buzzwords. But yeah, yeah. Um as well, questions about scale out, I think are so kind of it's kind of the number one you know, around real hybrid cloud scenarios and seeing Microsoft embrace that the other or another one is around actually kind of going down the stack a little bit as far as scaling
20:35
out the database engine, I think, right? Uh What's what's the context that you're going for a little bit here? Yeah, so that when we're chatting about kind of SQL server and storing data in different formats and you know, adjusting out the, the standard B tree with indexes. Yeah, keep going. Yeah. So one of the things that you also have seen in
20:52
Microsoft do is, you know, traditionally, since the sixties, we store data in databases in what's called a tree, which is a structure that allows me to get access to a thing uh in an efficient way. And that's great for certain types of data. And what we've seen is that kind of evolved over the years,
21:08
we've seen things like column store come out which changes the physical structure of the data. So I get access to larger swaths of data in a more efficient way, compresses it reorders it things like that. Now we're seeing in the kind of this hybrids center and I alluded to it a second ago with S3 connectivity is I can have data in different formats in different places and access it with
21:28
a thing called poly base. And poly base gives me the ability to essentially just access a piece of data where it physically lives, whether it be in parque format delta format CS V or many of the other ways that we can access data over poly base. And so that's what we're seeing coming out in 22 and SQL server 22 is the ability to get access to those data sets wherever they live.
21:47
And that kind of breaks down. I think the need for things like complex operations like ETL or extract, transform and load where I have to have these complex business processes to get data from format A to stick it into this database engine. So I can access it one. I could just leave it there and access it if I'm only hitting it periodically or I can
22:07
ingest it and stick it into my relational database. Um in a much simpler way before using just T SQL right to get access to that data, which I think is incredible. I feel like if, if anyone on here thinks about ETL and lives that extra transfer load, even if you're not going into newer ways of doing it,
22:25
pure S stuff that can help you there. That's a teaser, reach out to your local sales team, you know, they have good stuff to talk to you about with ETL. The other piece there though. And this is kind of finishing this one up is that um there is sometimes we don't, we think about maybe programming languages or scripting languages,
22:40
whether OK, let's go way back in the day. We, we talk Cobal. If we're gonna talk scripting, it's bash or maybe Pearl or PHP or, or newer stuff like Python and Ruby, et cetera. OK, Ruby on. But there's a whole language as it relates to sequel that Microsoft has continued to invest in. I'm curious what you've seen there and, and
22:57
maybe even actually the the, a little bit of definition first for the folks who don't know what T sequel is for T SQL. For sure. Yeah. Yeah. So T sequel is the way that I ask the database to do something, retrieve data, change its configuration. Um Essentially interoperate both uh from a development standpoint and an administrative
23:14
standpoint to ask the SQL server to do something for me. Conventionally, it's gonna be to put data in uh or pull data out. And what we've done over the years for the last 40 some odd years is we've written T sequel. The, the standard has evolved and it's become, you know, more powerful et cetera over time.
23:31
But the thing that we're seeing in Microsoft do whether it be all the way from Azure SQL edge, right? In IOT devices all the way into the cloud and into your on prem data center is it's the same T sequel language for me to interoperate with data wherever it physically lives. And again, that's a really another powerful construct in that I can start collecting data
23:52
and inter operating with data regardless of where it's being collected with tools and techniques that I'm already familiar with as, as, as a developer DB A, right? So I can run T sequel on a IOT device uh in the middle of a field or in the middle of the ocean, right? When it comes back in a range, I can track the data and,
24:09
and that's I think an, an extraordinary technique because that's the one guarantee that we've had over the years. Other than the fact that data continues to grow is that languages are gonna change, let it CC is never going away. But I have feelings about that anyway. Or maybe it's, yeah, all the rust people just came out and added me on Twitter.
24:30
We haven't done developer and language focused ones. We have to figure out if we can make that a coffee break to still link it back what pure does in some way. So the, the, the, the, the, the theme I love there is that that for better or worse businesses have investments and business processes and those things are tied to languages like insert personal memory of,
24:46
of an organization trying to get off mainframe for years because of all the stuff you had to rewrite. So that investment in t sequel helps take years of business processes and extend it out so much more. So that's like it's maybe down, down in the weeds a little bit, but it's serious value from a customer standpoint kind of thing.
25:03
Yes. So with that, I'm gonna end up in the first poll here and I will go ahead and share it back and we'll launch the second poll after that. If we learn anything from the first couple months of doing polls is to give folks time to listen to, to chime in. So actually seeing this, we've got a decent split here, I think between storage focused,
25:21
database focused and actually equal focus. Only a couple of folks saying they're just here for the coffee card. So, you know, I appreciate the honesty. I'm glad to see it's not too high kind of thing. Um Man, the, the highest one by far is 2000 or 2005, Anthony for sequel server version. That's just, that's fun.
25:40
Yeah, 2005 was a, was a huge release where server like really became a platform to build something on and it just, it got better. That was a huge release. Like uconn was the code name. And after that, it was, it was really the foundation I think of where sequel Surfer is today. Like a lot of bits and pieces you put together.
25:58
Then then finally, um 40% of you all have a DB A team, not just a DB A, that's number two. And in the interest of time, we're gonna stop sharing that one. And actually, uh Emily, I'll go ahead and let you in mind launch poll number two. because I think we'll need to make a little time.
26:14
This is just like it's too much fun. So then we just get going because it's good stuff. So uh the questions here is you see when you expect to start using SQL server 22 because that's what we're gonna talk about here. And when you do your next install where is it going to be installed? Because there's a lot of options there too. So actually I should have clicked over here to
26:30
poll number two. Hey, there's the slide for it. But one actually is we're gonna start to move into a little bit of pure and sequel server stuff. Although not entirely what I find fascinating. And you and I were even chatting about this, like S3 is not necessarily a Microsoft originated thing, you know, so and not invented here syndrome.
26:51
Sometimes you see in in right? But Microsoft and sequel server 22 is really embracing S3, which I just find fascinating, especially from a question of throughput and parallelization. Do you mind walking through some of that both from a Microsoft standpoint and a peer standpoint? Sure. Sure. So what we see one of the big,
27:11
I think there's a lot of stuff in 22. Um but I've spent a lot of my time focused on two key areas, what's happening in with this, what's called the logic integration and just internal engine enhancements, right? How data can move around? And one of the things that Microsoft realized is that, you know, you have to be where the data is and a lot of data is in S3,
27:33
it's an object, right? And kind of the idea of that hybrid cloud thing that we talked about a little bit ago and also this idea of being a kind of a data hub, right? SQL server can just go get the things that it needs when it needs the thing, right, that piece of data, whether it lives in some other file or whatever or et cetera.
27:51
Uh One of the other things that Microsoft saw as a lot of customers and I saw this in my customer base too, in consulting is they were backing up data and then syncing into S3 because it was um especially especially in a, was like a cheap and deep in the cloud itself. Now pivoting it back towards what we see from an operational standpoint. We talk about backup and restore S3 is the idea when we're using things um like on prem
28:14
scenarios, we can definitely add value in terms of throughput and also concurrency, right? And I'm sure a lot of there's a fistful of DB A S in the party and the call today that have a giant database that's hard to back up, right? I'm sure there's a lot of us out there and I'm sure there's a lot of us out that have lots of databases that are hard to back up,
28:34
right? Because I have to figure out how to squeeze all of these things in my backup window, right, which could be this big depending on how your businesses recovery objectives. And so that's one of the scenarios on prem that this S3 integration brings to the party is that we can drive concurrency in backups in a much simpler way than we could in the four times.
28:55
We can also run lots of backups concurrently inside of S3 or um the S3 into a backup target. And our favorite S3 backup target in our world is flash, right? Was always crazy. Here is like when we created flash blade. But so we're going back 67 years. I don't,
29:15
there was no visibility that Microsoft in 2022 was going to do crazy things. So that's three. So it's almost like this perfect alignment of the planets or whatever another analogy you want. I'm like we've got an amazing and fast S3 stack and now Microsoft wants to consume that with single server. Look at that back to you. Yeah.
29:31
So like the core thing, this comes down to real concurrency inside of TCP. This is where this kind of unlocks a little bit is that we can run that concurrency at the network level and drive higher throughput rates and drive higher concurrency rates. So if you have a large environment, this is a big deal, this attacks a big problem that you might have because this idea of,
29:50
you know, and some of the customers I've worked with, they're backup servers, servers, plural, right? Can add to the complexity of the environment because I have backups kind of sprinkled across different backup servers now. But with something like flash blade, I can stick these on a flash blade and have the
30:04
access to them in a single name space, but then also replicate them somewhere else, like, you know, checking the check box to send it to another S3 target, whether it be another flash blade or in, into a A BS itself, right? This is pretty cool stuff because you have one job as a DB A job one and job one A is one is take the backup and two is get it out of your
30:21
data center, right? So, and we help that problem there, there's also the extension and, and you were actually alluding to this a little bit too in the trend. So we're just gonna continuing to kind of dive deeper about enhancing the sequel server engine to handle object storage. And that's just, that actually blows my mind a
30:40
little bit. Although for me, I go back a little bit to like my sequel and the different uh different engines that you could have in there, it's not quite the same thing, but it reminds me of that a little bit. Yeah, in the sense that I can get access to the data and we talk about wherever it is in the format that it's in because developers are gonna produce data systems are gonna produce
30:59
data in, in, in a particular type format. And so now we're seeing this kind of this hybrid ness come together where I could just go get the data on demand regard, not regardless but starting to become regardless of the kind of format it's in, right? And we'll see that, you know, that I'm, I'm gonna guess that that list of bio types will get bigger over time.
31:18
Arcade Delta CS V, et cetera, right? Um To get access to that data. Right. Cool. So in case you didn't see it already, you can just go to not tina dot com, but Anthony's not just talking about this for today. I made sure to put the URL there. There's another one to his blog where actually he's been talking and blogging about this stuff and actually testing it in labs.
31:36
And I think you were earlier today, you were talking about driving even over like over the internet, like three gig of throughput with some of this kind of stuff, some, some really fun numbers to talk about, you know, kind of thing. So doing this hands on too, super big fun stuff there.
31:49
So poll number two, let me go ahead and end the poll here and I'll share it back with everybody. So everyone can see uh when do you expect to start using sequel server? 22? Um Man, this is just a real mix here between the top three actually. So I think the uh the year after G A first service pack that was your preferred, you know, in a DB A for a while with some scars to show
32:12
it kind of thing. Answer, right. Yeah. And then um so windows and VM Ware. So you're gonna say, oh, I know. I know. Yeah, the modern, the modern service release cycle no longer has service packs. That's what the Winky Emoji there is for.
32:26
So, so when for the next version of SQL server ware to put it, this was a multi select, you know, select all. Um This is, this is fascinating to me that VM Ware continues to be the top choice. Um And not fascinating, like I don't expect it, but there's turbulence in the market and people are doing things and there's all this cloud stuff going on and different trends so far and
32:47
away. Windows. VM Ware only is, is the option but there's a healthy showing uh for some of the other ones there too. Yeah. Good to see you. Good to see some Lennox out there. What's up, which is just fun given growing up in the days when those two things were like this,
33:04
but now they're not anymore. So that's cool, Emily, if you don't mind launching poll number three before we dive into our last section. Uh So this is as well. Uh This is a little bit of a shorter one. So how might you use sequel server three capabilities?
33:19
We intentionally didn't want to ask this question till Anthony just finished walking through what he went through because otherwise you'd be like. So, so what's that? And then also um as a leading into our next section, uh are you using snapshots for database refresh because we got a little bit gutsy with the final uh with the final name here, the bending space and time,
33:39
not just snapshots, you know, jumping in the time machine with Argenis going back last year when you talked about this or if I'm just gonna go. So II I still try and stay relatively technically deep, but at times, I'll just go into like I, I'll, I'll consciously be a hype man. So volume snapshots for databases is literally a superpower.
33:55
If you want to be a superhero in your organization, you figure out how to use this for instantaneously synthetically creating huge amounts of excess extra data. You know, it's a snapshot, it's just meditating and pointers and bending space and time. OK? I, I feel like I shouldn't, I shouldn't quit my day job to be a hype man because it didn't quite come out,
34:14
right? But there's, there's a lot more depth there if you don't mind taking it and continuing to run with that. Sure. Sure. So volume snapshots to get access to database data, right? I see some comments in the in the chat about different scenarios.
34:28
Um And I'll, I'll try to address a couple of those. So when I, when I talk to customers, I say if you're copying data or backing up and restoring data or moving rows between server A and server B, you can stop doing that, right? And some of the scenarios that we cover are the ones that we have here like DEV test refreshes.
34:48
Um You know, if I ever have to do a backup and restore for data protection or, you know, be before an application upgrade, we can live leverage snapshots to help us in these scenarios. So imagine like being able, you know, having to file the change management request to do your big application upgrade on Saturday and you pull, roll back time, five minutes, right? Not 7.5 hours or whatever it is to do a full
35:07
database restore. That's gonna be a huge boost to the confidence of your ability to get the system back and functioning. If there is an issue, you can leverage snapshots, literally just revert back in time, right? Similarly, if I want to get data between instances, I can take a snapshot and clone that volume to another instance to get access to the
35:25
data in that thing. And that could be a viva that can be a V MD K on VMFS, that can be a physical RDM or a real server, right? We have code samples on code dot pure storage dot com that go through all of those scenarios, right? That's huge, huge, huge there. So if you want to bump to the next batch of slides or um bullet points,
35:44
there we go. Yep. Yeah, it's because what, what, what we're doing here is if I have to move data between A and B I have to spend compute networking storage and all the overhead associated that blowing out my buffer pool. If I have to pull do row level operations, I can make that go away. Right. And, and leverage the power of the storage
36:02
array, you get that data between database instances. And this has been a core technique of flash ray for a long time. But you know, I'm gonna, I'm gonna tease on, on the, on the poll results a little bit. It's about uh 56 to 44 the distribution of folks that are using that aren't. Yeah. And so I'd love to you.
36:19
We can spend a little more time getting deeper into that. Probably not today, but I just want to talk about the capabilities there and how we can get that type of functionality into your platform because restores are hard, right? Whether it be for an actual recovery operation or just getting access to the data for non broad testing and things like that,
36:36
right. The one thing I think of here is that sometimes I have to hopefully in kind of gentle educational mode without being a jerk for sometimes I've worked with a lot of other platforms that have stored snapshots. And historically, there were a lot of benefits whether it was copy on first, right, or performance stuff or maximum you can make or the snapshot was chained to the
36:53
original volume, there was a dependency or you delete snapshots and then a a waffle iron or other cleanup process will kick off that have random performance impact. So all this stuff that goes in there. So there's almost a part of me, it says like I almost want us to not call them snapshots is pure, like call them some other fancy, cool sounding name.
37:08
So it's not associated with all the sins and problems of the past, you know, kind of thing. So, yeah, and I see a couple of specific scenarios being discussed in the chat. Um Just please feel free. I will, I'll throw out my contact information uh at the end. Um And I will talk, you know, we can break through those scenarios if you're,
37:28
if you, if you have this blend of technology, I saw FC I question roll by. That's literally part of my job is to help unblock folks in that case. So please feel free to reach out, put that in the Q and A as well. Uh Melody and Zain are typing away and we'll have a specific time for Q and A and we'll try and actually hit some of those too if they're in the Q and A section,
37:45
the other piece here. And this is almost uh we're coming down the home stretch, but actually, amazingly, we're kind of on time is there's, there's some new database engine features and this is where it's almost, this is cutting edge stuff. It's a little bit of feeding your inner technical soul or geek or whatever we want to call ourselves these days kind of thing.
38:03
So, qat offload is really neat. You want to unpack it then? For sure. So we know flash blade can move buckets of data really, really fast, right? We can, we can feed data into it really quickly and we pulled it out of it really quickly. But what if you're compressing that backup on a production sequel server instance with lots,
38:21
lots of cores and lots of lot to work with. Let's say your sequel servers are running at 60% CPU and you're OK with that, that's a comfy for your general workload. But then you kick off a backup with compression. What happens? Lots of things go red, right? Because you're spending all that time getting that data uh out of the database compressed and written into the network,
38:40
right? What it, what the Qat card does is it takes that compression capability which generally be done on the CPU and puts it into a special card that's built for just that. Uh And so that in the performance testing that we've done internally with workloads running and without loads, workloads running with compression, you'll get a 1 to 2% CPU bump,
38:59
which if you have a very large database, this is huge. If you have a 24 7 shot, this is huge because now you can perform database maintenance operations like a backup without um the CPU overhead associated with that, right? And that gets to become a huge deal. And we've noticed in some of our preliminary testing internally,
39:18
we're yielding about a 10% smaller backup file using this technology because it's a different compression algorithm inside of the um card itself. Now, your MS may vary. This is a specific data lab scenario, but we're using this specific QAT technology on SQL server 22 and getting really, really good results and less CPU burn and smaller backup files.
39:39
It's really cool. The other piece there from kind of a looking forward standpoint. Hey, we, so we started with T sequel or not started but you know, but there's even more there too cool. Cool. So, all right, we'll put it into the chat or like an ad hoc Q and a who's had a reseated availability group and it took a long time,
39:58
right? So this is a really neat capability. It T SQL based snapshots. Um This, the nexus of this technology was an evolution of VSS because VSS was very baked into the operating system of Windows and VSS is what gave you the ability to take what's called application because it snapshots of um of a database, right? And it would create the database.
40:22
You could take a snap of the volumes and you could do something with that thing. Well, well, it came out a couple of years ago, sequel server on Linux, right? Or could you put vss onto Linux I don't know. Probably not so much. Yeah. Yeah. And so the sequel, server engineering team wanted to have,
40:37
you know, feature parody across their platforms and they couldn't bring vss over to Linux. And so what they did is they pushed up into the database engine, the ability for the database to be put into what's called snapshot mode. And then I could take a snapshot of the volumes supporting that database. Now, the key difference between a volume based snapshot and an application consistent snapshot
40:57
to a DB A is this, I could bring a snapshot back and put it in recovery mode, which means I could then restore diffs and restore locks. This is a huge capability for fast recovery of a database because if I have a giant 40 terabyte database and I wanna get transaction level recovery, I could snap back instantaneously restore a couple of logs and present that back to the
41:18
business to consume. That's huge. Right? I'll snapshots our backup. It just really fast restore, but we can get into all that if you want to orchestrated. Yeah. Yeah. Yeah. All good. But now I have this thing that I can recover very, very quickly.
41:32
That's point in time recovery. But one of the capabilities that we couldn't do in the before times the VSS uh snapshots is this, I couldn't see the availability group or build an A G from snapshot. And now you can, so that's scenario where I have a very large database. I could build an availability group across two replicas across two different arrays nearly
41:53
instantaneously and almost independent of the size of data, which is crazy to think about. So when I create the availability the first time, yes, I have to move the data between two different arrays, but that's going to be data reduced. And generally it's about 3.5 to 1 on pure technology, right? So 40 terabytes now 10 terabytes gotta move.
42:09
So let's say I'm up and running, right? And for whatever reason, I have to receded the secondary on that target using this technique at the blog post here. If the majority of that data is already in a target array, when I send that snapshot over, only the incremental changes will have. Which means, yeah, which means I can bring my business back into an H A posture significantly
42:30
faster than I ever could have before because I had to literally move the bytes and bits of bits between the two sequel server instances with either backup restore or direct seating. This is I think such a huge, huge, huge capability in 22. It was the first thing I blogged about once I was able to actually blog them because I think this is gonna be a game changing piece of talk, right?
42:49
So as noted, you said it, we're gonna highlight it even if the even if you it's hard to type out the whole length there, just go to nono dot com, scroll down and post, you'll find it a little bit. It's super cool. You've actually, we're talking about it and then you've got the full walkthrough poll number three, 1195 responses, which is awesome. So, I'm gonna share that back and then we'll
43:06
launch the last poll because we're just about at the home stretch. Bear with us. We're gonna go a minute or two past the 45 minute mark before we jump into Q and A. But I think that's all with everybody here. Hopefully. So you were watching this live Anthony. I'm gonna to do question number two. So that actually edged up to 61%.
43:21
That makes me happy. I think that makes you happy. There's still a lot of opportunity there. More cool stuff you could do for, you know, uh about two, about a third and, uh, I, I'll actually just any, what was your gonna be your guess about uh sequel server S3? And how that, that one was gonna break down the results there?
43:40
I think the both is skewing the results a little bit. I don't know if that's like an additive thing. But, um, the, that's the folks that aren't interested. I, I understand but I, I like that the adoption is nearly, I guess you could say, let's say 89%.
43:53
That's pretty awesome. Very high about it. Thinking about doing it. Yeah. Versus like uh, it's a new feature, I'll come back to it in five years. So, um, OK, last poll and then we'll call it a home stretch. So, um, Emily, if you don't mind launching the last poll,
44:05
because this was really, we wanted to do this one more polls, I realize this time, but each of them were kind of linked to the sections a little bit. So curious how you see yourself using T SQL based snapshots. I see the answers there and then is S3 object integration useful in your environment. Maybe we could have put that with the previous one.
44:23
Maybe not. I'm gonna leave that pull up for a moment, Anthony, but we have um we've wandered through a lot here as always the time just kind of disappears. You know, you do good preparation, but then you kind of relax and you have fun. Hopefully for you too. Thank you for being um such an amazing guest.
44:40
I'm gonna jump to the drawing. Uh Any last uh any last sound bites you want to put in before we kind of finish the content and then we'll, we'll hang around for Q and A but any last words to take us home. Uh Honestly, I appreciate everyone's time. Uh And being like you, you can tell, I get a little excited about tech. I, I love being able to work with these things
44:58
and talk to fantastic customers and my coworkers that are on the call. Um I'm just really, really happy to be here. So thank you for everyone for being a part of this. This is one of the Higher Tents we've had. Sequel is always a popular topic for good reason and sequel and pure. And that's why you're here for crying out loud, right, kind of thing.
45:14
So to the drawing, I mean, we got enough folks that I don't think everyone stayed around just for the drawing, but probably some folks. So to Brenda W from Texas, thank you so much for joining us. You are the winner of an ember mug retail value 100 and $30 the kind you can control with your phone just because it's cool, but it's actually kind of useful too, you know. So there's a little bit of like because we can.
45:33
Uh but actually my, my, the, the one that I have, my wife uses, she drinks a lot more tea than I do coffee and actually keeping tea at a constant temperature. It's all cool. Please make sure to join us next month for how sick is your data platform. And we're gonna explore how pure helps health care organizations and more and some finance with Josh Gluck.
45:50
I'm super, I've really enjoyed all my conversations with Josh. He's very approachable down to down to earth kind of guy, but at the same time has some fascinating customer it, executive experience with that. We are now. Well, we're like a minute and a half past. So we are now into the Q and A but before I do
46:08
that, um, we are, we're already a good number of folks who've done the poll. So, if you didn't get it in, you've got another 54321. I'll end that and share it back out so we can jump into the Q and A. So it looks like, uh, how do you see yourself using T sequel based snapshots? Uh All of the above is the winner by one, you know.
46:31
So, hey, that's kind of cool. Um None of the above, not too much, but then there's some individual use case interest, which is, which is interesting. So um any comment there, Anthony before you, I've had the receded availability groups, no, one thing I want to call out specifically if you're reseting an A G A common scenario is
46:50
if you failed over to Dr and you fail back and maybe that was an unplanned fail over in AYC mode, that means you got to get all the data back leveraging the techniques that we talked about here. If the majority of the data is there, then that receding operation becomes a differential leveraging storage based snapshots. And that's, that's a huge game changer for one testing and two,
47:09
the actual Dr failure getting back into not only um an H A posture but AD R A better Dr posture more efficiently and quickly to your business. And so that, that capability I can't under understate the significance of that. If you've been doing this for real, there's pain to that you can avoid in the future I think is the one had to take, um, one of the funnier conversations I've had to go back to
47:34
story time real quick. Um And actually I'm, I'm gonna pull the music up here a little bit quietly. If it's hard for folks to hear. Let me know, but we're moving into Q and A which is the relaxed, we talk about stuff. So keep going with the story. So I had a customer multi terabyte database and consulting.
47:47
Um the cio fought for and provisioned a 250 meg lane, right? But due to the nature of availability groups, the maximum throughput they would get on that link was about seven megabits per second, which is not 250. And so there was some hard conversation that had to be had around giving data between prod and dr because of that exact thing.
48:07
Um So we had to do some special stuff to get. I think it's like 20 some odd terabytes between the sites. If I just used um regular direct seating or backup restore, it was like a 32 day operation. Not great, not great. So gonna jump to Q and A um Anthony uh uh sorry Zain and especially melody. Thank you so much.
48:28
There's a couple that haven't been shared. So I'm I'm gonna take this super easy one. Oliver is asking for the links for next month's webinar. We actually had a little bit of a delay there. You will see that in the follow up email that you'll get so you can sign up for that one. So don't worry, you'll see that, but just didn't have it ready today.
48:45
The other super easy one is uh can you share Anthony's contact slide again? Well, yes, I can. That's just a click and presenter view, you know, kind of thing. But um you, you've got a wonderfully unique last name which kind of helps out on the Twitter side and the blog side and all that too. So it's actually fairly phonetic. I didn't have trouble spelling it when we met.
49:04
So that's kind of neat to look at that. Uh So there is uh one question from anonymous, I think we buy in the chat Zain. Highlight of this one. This one, I'm gonna toss to you Anthony. So how does Qat interact with third party backup tools such as red gate light speed, et cetera? Sure. So in the docs, you'll see um that,
49:26
that you need the card or not, there's actually a hardware mode and a software mode which is actually quite intuitive when you think about it. Because if I back up on QAT and I want to restore it to something that's not qat. Well, I need to be able to emulate the backup. And so you have software mode um to get access to those 23 different things. Now, the pop up to the third party question is
49:45
all that syntax is, is, is, is exposed as t sequel. So your third party application vendors are going to have to integrate that into their product. So I know a lot of folks use uh extended store procedures and things like that or the direct VD I interface. And so they'll have to just make that part of their feature set if that's something that they
50:04
want to do. Right. Very cool. Um This is just a human interest one, but it's easy. He's um and hopefully this is an awkward to talk about. I was asking if you finished your phd. I've had friends that have finished other ones that didn't like the dissertation and defense.
50:19
That's hard work. So, you know, that's the point on that. I have to do my dissertation in defense. Um But life kids, you know, all that stuff, but it's still, it's on the radar. So for, for what it's worth, actually, I think you can join. Um If I go back to the coffee break that we did with Merola Klasky,
50:37
we're trying to find that on here if I can find it. There we go. Um He's actually, uh he's also actually part way through his phd too. So maybe you and he can swap stories or commiserate or you know, or, or beat your chest because like, hey, look at me, you know, because like, like already done a lot of work on it.
50:53
Ok. Next one, this is kind of a fun one. It's a little bit open ended. So I'm gonna see where you want to take it. Best practices for storage of snapshots, for a software DEV company. My brain goes a lot of directions, but I, I'm gonna let you take first crack at it if you don't mind. Uh Let's see, best practices.
51:11
I know there's a couple of scenarios that are got um one of the things that when people ask me like what's, you know, an Achilles heel of storage based snapshots. The biggest one is if you have a long running transaction and take a snap when that, when you bring that database back online, it's got to roll back. And that's generally, that's a function of just the relational database engine more than
51:30
anything. Um But when it comes to the best practices, it's, it's more about the operations of doing it and writing good solid code to support what you want to do. Um More so than is there, are there any configurable or turns to, to make that better? There's really not. Um because, you know, in flash ray snapshots just work all of the time,
51:51
right? So I don't know if you wanna add to that Andrew. I'm saying about, I mean, in, in some cases, I with previous systems in maybe a different direction, you actually had to set up a specific space for snapshots and you exceeded that. They didn't work. And I'm going to more of a storage layer, you had to carve it off or if you hit certain
52:07
maximums, uh it would be a problem or if you wanted to clone the test DEV. Now there was a dependency on the source volumes so you couldn't clean up some stuff in production for a long time because the test DEV stuff was long running longer than you thought. So a lot of that sometimes when I think about it, actually, I'm just going into the storage layer, a lot of the kind of the requirements for how you'd have
52:26
to architect for software development at the storage layer for snapshots, that stuff just doesn't apply on here. It's just like it, it doesn't matter a lot of the limitations that I've seen in the past. So that's where my brain goes. Yeah, that, yeah, that's the, yeah, kind of lit up a couple of things in my brain too.
52:41
The idea is the snap is, is metadata. We don't have to track it over time with regards to like differential and things like that. But also the ability to move it or copy it somewhere else is huge too. So we have customers that might not want it running on pro but I can send that snapshot to another array and do my stuff over there, right?
53:02
One other one unusual would kind of go kind of sort of up to the army, but there was one that melody put a great answer to that. I want to make sure to get. Um There we go. So question from Chris. Thank you Chris for the, as always, thanks everyone for the questions.
53:16
What are the pure standards for building out balls for SQL or sequel melody already? put a great link in around um actually a pure validated design, a whole paper that we have around this around sequel server and high availability. II. I love the vials question, especially sequel server because the portability and the granularity that you get as it relates to
53:36
databases. I'm gonna make myself stop there and let you keep going. So I talk to customers a lot about that such that I I've blogged about it too. So if you go on to dot com, you'll see a VM Ware best practices blog that talks about the considerations, uh data file layout uh and things like that around um accessing data in vals,
53:58
but also the performance elements of that. And then link to Cody's blog, uh Cody Hosterman, who is one of our VM Ware experts that goes into a deeper dive about configuration and you know, the performance and considerations around that. And so looking at that, those two bodies of work hopefully gets you what you're looking for there. Very cool.
54:18
Let me do, uh, we're gonna do one more scan. We still got as always, it's pretty amazing how many folks we have just kind of keep hanging out with us and it's fun too. And frankly, so actually there's still some open questions. Let me go back and look over there. Um, I think we got that one. Hm. Ok. So actually
54:37
this would be, this would be a good one. So, um, from a Neil if I got your name, right? I think we'll end with this one unless you see any other there, Anthony that you wanna like. Oh, I want to answer that question. Cool is um I recently proposed sequel data compression. Our storage admin says this was defeat this compression technique applied by pure.
54:54
What's the advice? This is the bigger, this is one of the bigger it depends. Um There's two, there's, there's two ways to look at this, the storage admins way and the DB A S way and, and basically, if it makes your workload run better, then do it. Um Meaning because I can, you know the page is gonna be compressed,
55:12
it'd be compressed in the buffer pool. So I could store more data in the buffer pool at the expense of the page being compressed inside the array. Yes, your data reduction number is gonna go down, but you're literally storing less stuff because the page is smaller, right? And so it really depends the other key difference between storage based compression and um SQL server based compression is you can,
55:34
there's a, you can the vast majority of objects inside a SQL server can be compressed, but there's some that aren't inside a flash array. We treat everything equally, compress anything. So you'll find also that you don't need to go and have configuration everywhere to enable and compression if you do it inside the array. So it, it's really a balance. If it makes your workload faster,
55:55
then do it. If you see your DRR number go down inside your array, that's OK. Your data reduction because you're literally storing less stuff, right? I lied. I said it'll be the last one, but there's one other fun one that just, I know it's fun and I will, you're out there.
56:11
Thank you for asking this question. We didn't pay him to ask it. So the question is short, but we're gonna unpack it a lot. Application consistent versus crash consistent. There's a post on your blog which I, I believe and I want to just give due credit because you're hosting all this stuff. Now. It's one that our gist wrote and I think you
56:25
worked with them on at the time. To be honest about doing basically um 100 databases with crash consistent snapshots in the pure array and every single one passed check DB, it was like a torture test kind of thing. So there's part of it from a pure standpoint of the answer is yes, it doesn't matter. But I know there's more there.
56:45
Keep going cool. All right. So crash, consistent backup. The primary function of a relational database engine is to bring your database back online in the, even in the event of a failure. So this thing called the right ahead logging portal called guarantees that the state of what's stored in the data and log files will bring you a recoverable database,
57:06
right? Because the idea is if I walk into my server room and trip over the power cord and pull the power cord out, the data has to be consistent when I bring it back. And so if I come along and just take a storage based snapshot right now of the database and the data file, the database will be able to recover.
57:23
Right? And now I talked about a edge case a little bit ago, long running transactions, we have to roll back, but that's ok, right. That's that would happen if you did a backup and restore at the same time too. Now, that's a crash consistent backup. What you can't get out of a crash consistent
57:36
backup is point in time recovery. So my RT O or RPO, I always mix that up is the snap. I can't get to the, I can't get to a transaction in the middle, right? So through that over here, if I want to get to a transaction in the middle with point in time recovery I have to use application consistent stabs.
57:52
And so before 22 that was VSS and a 22 forward. It's VSS and the T sequel stuff that I talk about. So I can take a snap. I could take a log, I could stop at during the restore, I can get what I want from a point in time recovery standpoint using app consistent snaps. Either way you're gonna get a recoverable database.
58:10
It's really about the granularity of what you're trying to do. And then also unlocking the more advanced scenarios about A G CD and things like that. So I think that brings us right to time. He brought it in like one minute under the buzzer. So, and I look at that II, I went and used while we were talking, I was totally listening. I promised,
58:28
I looked at the blog post, put it in the chat because it's just a fun one of like basically beating up databases in ways that isn't supposed to work, but it did and it does. So, you know, look at that as think with that. We are right at time. Thank you everyone so much for joining us and staying with us, Anthony. Thank you.
58:44
As always, I I'll look forward to seeing you next week in person and a bunch of peers who are with us today. So as always, please make sure to join us. I should go here to be good for next month's coffee. Break. You will see the sign up link in your email soon where I'll be chatting with Josh Gluck about so much more good stuff. As always from pure storage,
59:01
on behalf of Anthony Argentino. Thank you so much for joining us today. Have a great day. Thanks Steve.
  • DevOps
  • Microsoft SQL & Exchange Solutions
  • Coffee Break
  • Microsoft Solutions
  • Webinar
  • FlashArray//X

Andrew Miller

Lead Principal Technologist, Pure Storage

Anthony Nocentino

Principal Field Solutions Architect, Pure Storage

What’s New with MS SQL Server & Pure Storage

Who knew that the best coffee break conversations would end up happening online? Each month, Pure’s Coffee Break series invites experts in technology and business to chat about the themes driving today’s IT agenda - much more ‘podcast’ than ‘webinar’. This is no training session—it’s a freewheeling conversation that’s as fun as it is informative and the perfect way to break up your day. While we’ll wander into Pure technology, our goal is to educate rather than sell.

This month, host Andrew Miller will invite SQL DBA and Microsoft MVP Anthony Nocentino to the virtual break room to discuss all things SQL Server landscape, SQL Server 2022, and Pure Storage. 

Our 45-minute chat will cover:

  • Anthony’s history as an independent SQL consultant, conference organizer, Pluralsight course author, & why he’s now at Pure.
  • The SQL Server Landscape & new storage-related features in SQL Server 2022
  • How to use storage-based snapshots to restore a database, copy a database, and seed an Availability Group replica, all nearly instantaneously.
  • S3 & SQL Server - yes, they can go together and why you’d want to do so.
  • How Pure Storage makes SQL Server environments better

Join us for a great conversation whether you’re a DBA or just someone who wants to learn more about what the DBA’s in your life care about.

01/2024
Pure Storage FlashArray//X | Data Sheet
FlashArray//X provides unified block and file storage with enterprise performance, reliability, and availability to power your critical business services.
Data Sheet
5 pages
Continue Watching
We hope you found this preview valuable. To continue watching this video please provide your information below.
This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.
Your Browser Is No Longer Supported!

Older browsers often represent security risks. In order to deliver the best possible experience when using our site, please update to any of these latest browsers.