00:00
This session is an ask me anything on virtualized database edition. So that means that you were talking about virtualized databases of all kinds. So I have experts here from pure who specialize both in virtualization and many of the most popular databases as well. So the idea is you get the opportunity to ask database questions or virtualization questions
00:30
or a combination of both and get true real life answers from those specialists. And I want to give them a chance to introduce themselves first, I'll, I'll quickly introduce myself here on the bottom right corner. I'm Melody Zacharias. I'm director of solutions and strategy here at
00:53
pure storage. I'm a Microsoft MVP, which is a um most valuable professional. If you're not familiar with that designation, I'm also a Microsoft artie which is a regional director for Microsoft. I live in, in Canada. And so we'll go to Anthony. Next.
01:13
Hey, everybody. I'm Anthony Sin. I'm a principal field solution architect at pure storage. I focus on relational database systems and cloud, uh predominantly uh Azure and SQL server. My name is Andrew Sullivant. I'm a solutions manager with pure. My background is enterprise architecture which has involved me touching every aspect of
01:34
databases but also up the stacks while Anthony loves his relational systems. I love my document based databases. I love my key value databases and all of those other lovely things in between. OK. Yeah, Ron Eakins. And that's me, top left corner there, that good looking guy out there. Uh I've got to say it, no one else will uh field solution architect of subject matter
01:58
expert. And uh you said she's an MVP in the world. We have Aces and I'm very fortunate to have the distinction of being an or a director. Thank you. That's Mr Ace there. OK. All right. So I'm not from pure, I'm from VM Ware. Let me say that. And second, my last name I think is the longest,
02:19
last name among all these people, right? So I have 15 and I don't know Anthony about you, but probably have 11. So that's me. I'm the uh I'm a senior solution architect for VM Ware and I'm the oracle globally when it comes to pretty much everything or, and Tristan. So my name is Tristan Todd. I'm a field solution architect focusing on
02:39
platforms, primarily uh virtualization, primarily VM Ware. Um I'm an XVM Ware employee. I'm a V expert. I run the VM Ware user group in Portland, Oregon where I'm based. Um I'm standing on the side of the room here just because I don't know. I don't know. I, I'm not a database guy I'm not a DB a guy,
02:56
but it is one of the most popular workloads and most common workloads that I talk to customers about. Thanks to Anthony. Um And I have to ask before we kick off, like what's up with the profile photo and then you in real life as far as did you shave your head out of solidarity for me? Thank you. I, I just wanna thank you very kind.
03:18
So there you go. We have a whole bunch of nerdy people up front here just for you guys. Special for you guys because we wanna answer all of your questions. So make them as difficult as you can bring it. We want to hear from you.
03:36
These are all the different topics that you have experts here today on. Don't hold back. There are a lot of challenges in databases, whether it's volume managing its scale, looking at spend databases are growing all the time. The volume never stops.
04:07
It needs performance. Databases are always in need of more and more performance. No DB A ever said I need my database to go slower and you need up time. You need to be able to move that data to another location.
04:27
What if there's a disaster? What if you're worried about ransomware? What about the ownership? Visibility and what's happening in that stack? How are your DB A and your ad men getting along or are they still fighting over? Who has access to? What organizations handling shared data across
04:49
borders. What about compliance? How are you dealing with that? Is compliance an issue in your organization? None of these topics are off the table today. Data sovereignty localization, whether you're in the cloud or on premise are key issues for every organization today,
05:09
whether you're in Europe and you have issues with GDPR or you're here in the US and it's IP A all organizations have some sort of audit or compliance regulations in place. It can take a long time to set up a server and modern platforms are important in getting things set up really quickly. Pure is one of the best you heard Charlie in the keynote,
05:37
I think or no, maybe it was Sean talking in the keynote this morning about the little card and being, having, having a business card to be able to set up pure, that's really important to us. Simplicity is important. So even if your question is about how things are complex in your system and you want to know how to make things diff um sim simpler.
06:00
If there's an api that we can offer you or a way to make things simpler or you think something within your system is too complex. Ask a question about it, whatever it happens to be, I guarantee you there's probably somebody in the room who can make it simpler for you. So now we wanna hear from you, I want to hear from you and hear what your questions
06:25
are and I want them to be hard. I don't want these guys to have an easy time of it. Yes, my friend. So it's uh it's along what I talked to you earlier about. Yeah, Anthony. So the question I have and, and this is just the best practice from your experience and it'll probably be for both of you.
06:48
Um Setting up a fell, fell over cluster instance of sequel across two servers shared V MD DKS or RD MS who in the room is running uh RD MS in your vsphere environment. Anybody does anybody used to run RD MS? Is there anybody in this room that loves RD MS, right?
07:14
The concept of, yeah, Ron. Ron. Ron loves everything. I don't. So RD MS are not an antiquated technology and maybe you want to chime in on these topics as well, but it's a, it's not a deprecated technology, but it is kind of an archaic technology in that life, cycling RD MS um keeping things up to date non disruptively are very difficult to do.
07:36
Also data protecting um copying RD MS is very difficult and the problem is very easily size uh solved by moving to. So actually, um you know, easily kind of converting an RDM to A or implementing a val based shared disc across to um the virtual machines on E seven or above, correct? Because what, since it's a um you've got um you
08:03
know, more modern telemetry. You've got the ability to um do pure snapshots for that single V ball. Um It's a more, I don't wanna say future proof is that fair to say V balls are have a future. There's a future in. Definitely, definitely. Yeah, but I, but I would say that's a very mainstream use case is moving shared from RDM
08:24
to V balls in database environments. It's a, I don't want to say it's a softball use case for, but it's a very common one. So, so quick. So a quick comment. So you know, we have always been and everybody has been exposed to this Holy War, holy War, right? But we have, you know, to RDM or not to RDM, right.
08:40
VMF, you know RDM, correct. We always had this, this battle, the reason why and I'm customer facing as well. I'm partner facing as well. The reason why one would go with RDM more so over in the oracle land is well, I love my sand back up so much, my replication so much I can't live without it. I need to hug it every night.
08:58
That's my use case for the RDM, right? The other thing that people usually used to come up and say was noisy neighbor, right? When you have a lot of BMTK on the land, you don't have that, you know, you don't have the granularity to provide that kind of storage policies and so on and so forth. So you used to have that noisy neighbor where
09:14
we see BB is basically a crossover. You know, you're now you're able to cross the streams. It's basically a crossover between your V MD K and RD, which means you can have your cake and you can eat it too, right? So you have your granularity of A V which essentially is the storage under the V MD K.
09:31
But at the end of the day, it's an RDM for crying out loud, it's a sub, right? So to trust point, a A or an RDM can be converted as a V, right? And assigned to a virtual machine and low crystal that becomes a V MD K for you. So what's very important or what I see is very important. Let me ask you the question, right?
09:51
If you look at the life of a DB A oracle, my sequel sequel server, right? What do you think takes a lot of his or her time? Is it day one or is it day two, day one? Is installation configuration or day two? Is your backup restores, cloning, patching, refreshing migration, right? And all of that.
10:10
So my background, I've been, I've been doing oracle databases for 26 years. I was an oracle DB A before I joined VM Ware. So I have been through that pain point. I have been through the cycle basically makes all of these very easy, right? So there's, there's, there's a lot of use cases when you talk about and I actually gave a example to a customer and I'll shut up in like
10:31
30 seconds. Right? But I gave an example to a customer. Imagine you have a virtual machine and imagine you have three databases. And my use case is I want to take a backup of DB one. I don't want the backup of the virtual machine. I don't want the backup of DB two and DB three.
10:46
Tell me how you do it if you're on vmfs, right? Either you take a land based backup, which means I have the other garbage along with that, right? DB two and DB three or if I took a virtual machine backup, I pretty much get everything. But the, all I need to do is take a backup of those that is part of DB one and attach it to another
11:05
virtual machine. Hey, that's DB cloning for you. So I just gave you a use case here. So makes life of any DB A or any application owner. Very, very simple. Next question. Hey, um I have a, I, I'm asking for a recommendation here.
11:34
So I have a customer looking to virtualize finally their databases. So bare metal uh using volumes mounted uh through their storage, they're trying to go uh Let's put them on pure, let's go VM Ware. Uh This is a sequel 2019 standard. What's your recommendation? Just because uh I'm not a DB A and there's a
11:55
lot of options like always on uh cluster share V MD K. Like if if this was your environment and you're going from bare metal to virtualize, what would you recommend? Are you looking for performance availability? Yes, I think uh middle ground of both definitely uh yeah, closer.
12:19
So performance and availability will go down that road. So the the the case for was made, right? And so that means you're gonna get and this sounds not trivial, but it's super important. But the granularity of the observ ability into a gives me a lot of information as ad va right? It's like a performance team better.
12:37
I can find hotspots better. I can do the things that I want to do inside the database. So is definitely the first step there. Um There is some calculus around that with compatibility with other features and capabilities in our uh our world or uh if you're using any sort of um replication technologies like the and things of the house. So you do have to do some calculus on,
12:58
is it compatible to use a in the exact customer scenario that you're talking about? Um Because when you go vmfs, you know, it's got the longest history and the most compatibility across all these different technologies. And keep me honest there team VM Ware, right? Uh from an availability standpoint um Before I came here, I was an independent consultant for many, many years and I taught,
13:19
I made a lot of money deploying availability groups. And I made a lot of money doing training and all this fun stuff. And I taught a day long availability group workshop. And the first slide of that workshop was don't, right? Because you have to really establish what the recovery objectives are for the platform that
13:34
you're trying to build. And can you solve that problem with a simpler technology? Right? It might just be what's the VM ware thing where it keeps the VM online if the host fails, I think what's that called each other? Well, there's AJ DRS and then there's also continuous availability.
13:49
So the the the the idea of can I get the availability out of the platform without involving any special technologies? Right? So fail over clusters availability groups up levels the skills required to support that platform, right? And so if you're doing ad R scenario, do you want to have to phone the DB A S every time?
14:09
And so those are other things that go in the decision process on which technology that I choose to build a platform that's supportable, that lower right icon that was on the screen that said it takes X amount of time to deploy uh a new server that's actually an anti pattern to availability. A lot of customers that I worked with over time if I could deploy configure and get a sequel server up in five minutes and blow back all my
14:33
databases in 45 minutes and I have an hour sla I met my, my, my sl a without having to involve any special technologies. Right? And so I would try to go after that before I brought any harder tech to the party. So am I answering your question, sir? Yeah. Yeah. Cool.
14:49
Anything else you want to get into in that one? Are we? Good? Good. I can keep going all day. Had a lot of coffee. So I think this would be a general question is a snapshot, a backup for the record. He asked me that during my interview just
15:08
saying. So I, I'm off the hook on that one. Well, actually let me, let me grab that today because things have changed in that space and then I'll yield it to the floor here. Uh So SQL server 22 integrated, uh what they're called T SQL based snapshots or cross platform snapshots, which allows you to have uh the ability to take an application consistent snapshot without any
15:30
external software, right? I could put the database effectively in a backup mode to take a snap. Is that a backup? Not quite, but if I take that thing replicate it somewhere else and can restore that database starts to smell like back up a little bit, right? And I also get point in time recovery and all
15:45
the other things because that's all a backup really is that I take the bits and bytes, put them somewhere else and kind of get access to that data again. Right. So that's changed a little bit since 2021. Right. Yeah, I, I'll jump in that one as well. So snapshots are complimentary from an oracle point of view.
16:01
So, uh but you know, most oracle D rays will navigate towards an arm man backup. Uh It has some advantages, some disadvantages as well. So, uh yeah, use them both. Uh and it has changed a bit, you know, in the same way in the Microsoft world. So what we're seeing now is people taking story snapshot their database.
16:24
One of the disadvantages of soy snapshot is not data aware. Uh So there could be a block corruptions in your database that you don't pick up by doing a story snapshot. So if you are doing a story snapshot as your primary backup mechanism, you need to be validating your database periodically to make sure that there is no block corruptions. So from an oracle point of view,
16:43
if you're using R man, you're going to detect those. But if you're doing a story snapshot, we don't know what's going in there. So if there is a block corruption, the only time you'll see that block is when you access it through the database and you don't want to have your database with a corruption that's been in there for months or years because you haven't checked the integrity of it.
17:02
So DB verify or do an R MAN database validate to check the integrity database. Another interesting and we always think of blocks snapshots. Uh I'm working with a, a gaming company at the moment and uh they want to eliminate backup time completely. So what they're doing is using their stand, they're having two standbys.
17:27
So a standby database which is doing its job for, for DR purposes. But on the second of stand by, they are, they can do an arm man back up from that. They want to do that. And that's really a common use case. So they take their production load off the R man back up of the product server, put it somewhere else. But what you can do as well is you could take a
17:45
fast system snapshot on the flash blade, for example. And you then got a dot snapshot directory. So they suspend the redo apply. They don't have to do a database backup. They don't have to do any recovery because R man's been applying the changes all the time. So it's, it's a up to date version of the database.
18:02
So zero backup time, it's current. It's the integrity is good because you had redo streaming, applied logs. So take a snapshot of that and then they registered a dot snapshot directory in the R man catalog. So they can recover at an object level out of that backup. So that's that is where backup snapshot,
18:22
they start to blend together now and it's not so black and white, I'm gonna pile on a little bit to that one. And interestingly enough, I'm gonna talk about a database that wasn't on the list. It's Hannah. Um So Sap Hannah is a weird database that is very purpose built for SAP environments. However, from the snapshot perspective,
18:39
Hannah was built in what I would call the modern era of databases. That's like the 2013 to today kind of world where they took the lessons of the last 40 years said we're going to unlearn them, we're gonna do it properly. One of the things they did was they actually implemented their engine in a way which was snapshot friendly. So application consistent and crash consistent
18:57
snapshots. It was very clear what they wanted to do, but the problem came in exactly the same way, Ron and Anthony brought it up, which was how do you know that when you've taken the snapshot, it's consistent and it will recover because the point of a backup is to be able to recover it at the end of the day. So in one of their revisions,
19:14
they released something that said, ok, if you've taken a snapshot application consistently, we've got a tool that will go and check every so often how um consistent it is and tell you if it's corrupt. And what this told me was, there was a trend of we really, really like snapshots for specific use cases and they seem to be popular with their customer
19:33
use ba you their customer base, but they realized that there was a hole in the pattern of going from taking a backup, ensuring it was consistent and then recovering it at a later date because if it came to recovering it, you couldn't recover it. Everybody screwed. Um My SQL started to play around with this with the way they were formatting binary logs.
19:55
Post gras also started to play around with this in very focused way with the way they formatted the writer head log. And what this tells us is if we start fast forwarding time, the elements are there to start doing consistency checks within the databases or to be able to rely on snapshots complemented by other mechanisms such as take a snapshot,
20:16
run the, run the logs forward with mysql or post or take the snapshot and then check it. In another way, there are actually very well documented ways in which to spin up a secondary post database, check what's going on and then drop it very quickly. So it's all, all the recipes are starting to be there. But I would say a snapshot on its own to answer box's question is absolutely not a backup,
20:40
a snapshot with a number of things built around it. Yes, it could be part of a backup strategy. Excellent. Do you have some another response to that Anthony? Yeah. So the use case, right? What are you trying to achieve with a snapshot? Get the data back.
21:01
Right. But there's another use case that customers that adopt snapshots and use as a primary recovery mechanism that I always tell them to still accompany that with native backups. Because of this scenario, maybe I don't need the revert production back, but maybe I need to go investigate a data anomaly for some business process that isn't giving me the right answer anymore.
21:20
And I need to go do some forensics on my data. And so I still, I would still prefer to have native backups for that because it's a different pattern that I'm supporting, not primary recovery in a data loss event, but database data maintenance over time, right? That's another reason why I still encourage folks to have native backups, get them on the secondary storage and get them out of the data
21:39
center because it's more of a archiving play or maybe even a regulatory compliance play than it is immediate recovery, right? Two drive. So feel free to put up your hand if you have a question. So sorry, I got a long question. Uh So actually a recent one year, I have a techno technological partners that pushed to us
22:05
uh new clusters and we run Hyper V. Uh So actually one of the applications that we uh installed is uh sharepoint, Microsoft, SharePoint. Uh The solution is all on prem uh Then the customers start to use the share more as like a dropbox like you know, Google drive. So we start,
22:28
they start to push a lot of large data can be large image uh documents or other structure data. And every month we were faced with transaction lock full. So it cost a share point to uh having problem. So I was thinking uh because I also have flash array and uh flash blade sitting outside the
22:55
new clusters. Uh What do you think? Because behind share point, I realized that it's sequence and pe and customers are using the share point like a dropbox. And what do you think if I want to rearchitect potentially the share point to solve this transaction log issue?
23:13
Uh Can I you know make use of the flash blade or flash array to rearchitect the share point? But the share point will still runs on the new clusters but somehow to point the data stores or or some some mechanism sequel servers that can you know put the data into flash array of flash blade to solve the transaction log issue. So when you say the transaction log, is it filling up?
23:39
Are you hitting the two terabyte limit or are you hitting the volume size on the individual volume uh I think hit the volume size of the VM because the technology partners, they say they never expect the customers to use it like a you know box. So the problem happened uh on a monthly basis. So we are trying to see how best because one of the solutions that it just extend your VM.
24:07
Uh But I, I see the problem because it will impact the performance. So I guess if there's ways to, so there's two problems to solve. One potentially is changing user behavior. Two is you don't get to do that and you have to solve that from a configuration standpoint. And the easiest thing to do would be to add additional log files or expand the volume up to the two terabyte limit.
24:29
Um Because that's gonna be the hard boundary for sequel server and its log backup size for the active portion of the log. Uh If you have to expand beyond two terabytes, you can actually add additional transaction log files to support that workflow potentially. But again, it kind of goes back to the user behavior. Um There's two other capabilities that you
24:47
might be interested in doing some customers that have a very low RP OS for their databases will run um their log backups in a hard loop, not on the time interval, right? And so that might be a way for you to make little rocks out of big rocks, but you still have the active portion of the log problem if you're ingesting a large volume
25:07
of data. Uh I forget the name of the feature melody. I don't know if you remember this one, but it's recent in one of the more recent releases of SQL server where they'll have um they'll kick off a log back up. If the active portion of the log goes past a certain volume, I'm, I can't remember the name of that feature. Do you remember that?
25:23
Or barks, you know what I'm talking about? Yeah. What's that? No. Um, but basically the ability to kick off a log back up if there's a certain amount of change. Oh, no, never mind. It's, that's another technology. I'm conflating two things.
25:37
Um, but I would attack the problem hopefully by changing user behavior or just throw space at the problem is really the only way to go about that. I don't know a team. If you all have anything you wanna add, never underestimate the ability of users to do stupider things if you enable them to do stupid things. So by solving the, the, you know, the the log
26:01
size limit problem by letting them put bigger files, they'll figure out how to put bigger files. If, if it's images or videos, they're just gonna get, you know, higher density, you know, or they're gonna be longer videos, you know. Uh That's what, that's something I want to explore. Whether I can make use of a remote block store.
26:25
Let me point to flash blade or something. Be because uh my my users are the pay master. So I don't think you can change their behavior. Are you in the legal industry by chance? Sorry. Are you in the legal industry by chance? No, no, no. Uh in the aerospace and defense technology. OK. So uh not, not legal, not legal.
26:51
We don't have a legal issue. So, so the question is whether we can, I can rearchitect to put the data into flash play, sort of like I saw a server as something called block store. Yes, I think it's something to do with. Yes. So, but that's a SharePoint thing. So sharepoint would have to build that
27:10
integration for you to be able to consume that uh SQL server 22 has the ability to store table data external in S3 object or BLOB. Um So you'd have to rely on Microsoft to make that change inside of sharepoint to be able to consume that service, right? Because you don't con you wouldn't be able to control or change where that table lives without them.
27:32
Well, you could, but then they would hang up from you and you call it support when something goes wrong, right? You, you, we pay a premium but you never get a correct answer. Yeah. Uh I mean, that's a very frank from a customer experience. Yeah. But uh when I, I don't pay anything to pr but
27:48
the support is that good. So that's a different. Yeah. Yeah, thanks. Thanks a lot. So, the remote blocks. All right. Yeah. Hi. Um So the question is on um a highly transactional OLTP oracle
28:08
database where um the idea is to do the snapshot by putting the database in a backup mode. However, the concern would be that since the database is very, very transactional is that going to cause any issues or performance bottleneck on the oracle database platform?
28:29
No, no impact. So when we used to put the database in hot backup mode prior to 12, that was when we were no, this space and the concern was then that you could be in that position for, you know, fracturing mirrors or for hours while you're doing a back up. Uh When you do a flash snapshot on a pure platform, it's instant.
28:52
So you'd only actually be in the hot backup mode for a matter of seconds. So the amount of additional reader you'll be generating during that time is negligible. Uh The other challenge I would say would be since or 12 C. Uh You know, we haven't had to put the database in hot backup mode to do a story snapshot. Uh There is a story snapshot optimization or supports and that's an option as well.
29:16
So even not have to do that. And most of my customers for development tests never put the database in hot backup mode. But that's a, that's an implementation choice. You're saying 12 onwards, you don't need to put the data in a backup for production. So there is a really good white paper on the pure site.
29:32
You can talks about all the different reasons why you don't need to do this. So the feature you look for is story optimization. Uh but that is a licensed solution that allows you to recover from a time based or an SCM based recovery. But that is an additional chargeable item but just with a story snapshot,
29:54
yeah, we don't have to put that not back up mode because pure manages the right ordering and the date time stamps and all the transactions that we write. So yeah, we can use those. I'm happy to have a deeper discussion offline if you want. Thank you. So, a quick comment on that, right?
30:09
I mean, yes, you do have the option of putting the data is in a backup mode, correct? And as Ron said, the full block images are locked in the redo log file, right? The other way is if you do a crash consistent recovery or if you do a crash consistent snapshot from a storage perspective, I think A IA I was when the Metal note was released as
30:27
long as right order fertility is guaranteed, the snapshot ID is stored and any kind of fracturing of block is avoided by the storage. It's a viable solution. I mean oracle did that in oracle. A because there was a lot of clamor from the storage industry because this is what to say any time you want to take a backup, shut it down or do a hard backup,
30:47
people came back and said no can do right? Most of the storage including PE MC, that's when I was when that note was released, right. Saying a crash consistent backup is a viable backup. As long as those three points are taken care of right out snapshot ID. And these stories make sure that you don't have any blocks.
31:08
I don't know how the story does that. But yes, so we have to do some kind of magic. So that's another way of doing it. That doesn't mean you don't have to do the way that you just described now. But I'm saying there's another option. Uh there's another way to do it. And if you're running oracle on VM Ware, right? Which I guess you are VM Ware snapshots.
31:26
VM Ware clones, right? They are the guarantee right out of fertility. So if you do any kind of Srm DS R, yes it does. And there's a white paper and I'll forward you the link. I am the author of the paper. So we did this validation.
31:40
So the best, the best way to find it out is run any kind of workload. I mean, you don't have to take my word for it. Run any kind of workload, very intensive slob, hammer DV, swing bench or pick your shoes, right? And then do your stuff and then check the alert log file. We did, I did not see, we did not see any kind
31:58
of corruption. So that has been validated. So VM Ware snapshots and I'll say this again. VM Snapshots, VMWARE clones, uh VM ware site recovery, right. And B CD R which is VM Ware cloud disaster recovery. I know this is APR session, right? I'm trying to sell a bit of VM Ware here as
32:14
well, but these products will all guarantee right order fidelity. So if you want to have the discussion, please feel free. I'm more than happy to talk about. Hi with um databases and virtualization. At what point would you consider that? Uh It needs dedicated hardware versus being
32:35
virtualized. Grab that one for a second. So in my consulting days, I broke systems into a couple of buckets. Uh basically the tier one stuff or tier zero if you fancy tier 23 and so on, right? And, and the idea is from a virtual database, you're, you need to provide an SL A tier at work, right?
32:55
A performance SL A on top of all the other stuff. And if I have a, a tier one mission critical database, the last thing I wanted to ever have to do is wait to get access to hardware. And the only way you can guarantee that is by dedicating a host to that VM. And so if you have that condition where you want all the goodness of running a database on
33:14
VM ware, but you need the perf and you need to guarantee the perf that was the way that I could guarantee the perf, right? And then after that, I would then oversubscribe the hardware based off of performance data. Et cetera. So, you know, if I have a tier two workload, I could oversubscribe. Course. Right. And I would do that memory,
33:31
not so much heart and disgrace, not so much. But the idea is that's the only way in my mind to guarantee the performance of a workload. And I think, I think the concept of, of um setting resource guarantees, setting up resource groups, um setting up quality of service to kind of defend uh a key system. And it's, it's access to resource resources,
34:00
you know, increasing the share values like in a environment so that it has prioritization when you're, you know, over, over provisioning, but you have a run on the bank as far as like CPU memory. I mean, it depends on resources, there's ways that you can game the system to make sure that some V MS are more important than others. Sure. Yeah. OK.
34:19
So VM Ware expert, fantastic resource and globally recognized. And when I made the, the statement that I made specifically about running tier zero workloads on VM Ware, I've been doing that from a consultant's perspective where I can't guarantee that I have access to someone as awesome as this to be on the other side of the fence to configure all those nerd knobs in an appropriate way.
34:38
And so, yeah, I know this is kind of adding drama to the party. He didn't let me finish. So you could do resource groups and you could do shares and reservations and limits and you could play that game. But I used to have hair about the length of melody before I became a full time c admin running virtualization. And I pulled all my hair out because I,
35:01
I was short on hardware and I had to make sure that tier one and tier two workloads had resources. So I played the game where I used every mechanism in the hypervisor stack to defend or prioritize resources. I don't think it's a sustainable thing to do in a production environment. I wouldn't recommend it.
35:18
Um You know, I, I guess for some workloads depending on if you know, if the use case involves life safety, maybe it makes sense to dedicate hardware, you know, and maybe isolate, you know, you could still virtualize which gives you extensibility and portability and abstraction from, you know, tedious hardware things. But it could be that you,
35:35
you know, limit a single database, large database or, or critical database to a smaller piece of hardware or a small number of, of, of critical databases to a small cluster. I'm not a real big fan of, I've never been a fan of standalone vsphere host, like cluster clusters are a good thing in vsphere land.
35:55
Um Something something in plus one availability. I don't know. But II I just, I've seen people try to squeeze blood from a turnip through limits and resource guarantees and it's just, it's painful to watch because you're always gonna have a run on the bank at some point, keep your linkedin profile updated. If you do that. Christian and Anthony focused on the CPU and,
36:17
you know, ballooning and memory starvation and all the other bits you have to get right and balance from the or side. We still see lots of physical databases and a lot of that is not so much the technology, it's more the fear of or licensing and being able to guarantee that you're uh you're only using part of a machine that you and you're like having to the whole machine.
36:39
So VM Ware and some pure partners have done some great work around how to configure VMWARE so that you're compliant if you want to use part of the machine. Uh but you have to get it right. Otherwise you might, and that's why most customers would dedicate a whole machine to an oracle platform because if you don't get it right, you're paying a lot for your or licenses and uh maybe you've only buy
37:05
four CP licenses, you got 24 there, you could be on the hook for the other 20. So if you haven't got it 100% right from the oracle compliance and audit teams to add from the to add from the My SQL side. Um My QL is a database that's been developed for quite a while and it's got lots of bolt ons to it. One of the earliest bolt ons for the storage engine interface was.
37:27
How fast do you want your storage engine to run? So there are actual actually parameters that say um how many I OS do you want to allow this one instance to do? Um And what this speaks to is a relatively similar vein to what Ron is talking about of. If you optimize at the application layer, you optimize at the operating system layer and then at the virtual layer, you should be theoretically OK.
37:48
As long as you've taken into account everything architecturally um that needs to be done. Um But as Tristan said, there could always be a run on the bank. You can never guarantee 100% perfection. All right, I wanted to follow back up on you guys conversation with limitations replication. So you guys said databases you'd recommend putting on because you get a benefit of RDM.
38:14
So with that being the case, let's say you have a database that you need a active cluster for, for DR replication at the site because it is a mission critical database. How would you guys go about that? What type of support is coming out around that from those lines and from dr other aspects. So, so active cluster is essential to these.
38:35
Well, we don't support it are not supported yet, right? It's coming, you know, um I can't, I can't publicly state like what's the date? Um We're working very closely with VM Ware to kind of ship um active cluster support for as soon as we can. I would say that we are um There's a great deal of Q A um
38:58
effort under way right now to make sure that we get it right. Because if we ship active cluster and support and we screw it up, it's gonna be bad. So I would say right now, I, I don't think I'm violating any, you know, ND A with VM Ware. Um If I say that it's coming soon,
39:17
um we are going to, we are going to make sure that we scrutinize that we really Q A that pardon the expression Q A the hell out of it before we actually make it G A. So I think um I would expect, I mean, or maybe you can comment on like is there maybe, you know, there, there's a big conference in August in Vegas, like down the street, right? Explore. And I think that was a session before,
39:40
before this session and that was done by, that was an excellent session done by Jason from VM Ware. And uh I think that was Nelson Nelson. That was an awesome session, right? They spoke about, I don't know if you were there. They spoke about the active, active cluster on the uniform, non uniform. Really?
39:56
That really blew, blew everybody's mind, right? But the amount of work and Q A that has to go into it to point absolutely. Due diligence has to be done. Right? And, and that can you confirm, maybe, maybe you have to remain mute on this topic, but that's gonna be something where you're gonna have to run Vsphere eight, right? It's not gonna be something that you can
40:19
maintain. Vsphere seven update three, whatever I can, if I can quote Jason and if, if I get into trouble then he's in trouble. That's what he said. Right. The guy before me. yes, it's gonna be Vsphere eight. It won't be back to seven and 6.7. So it will be V eight.
40:35
Yeah. So if, if you're if you're still tender from your upgrade from Vsphere 6 to 7 and you want this get ready because you're going to eight that the good news is the upgrading from seven dot X to eight is I mean, I I got through the process and I'm I it didn't, it didn't hurt, didn't cause dry mouth. I didn't lose any sleep.
40:54
It was a smooth process. 6 to 7. Whole different experience. I was, I don't want to go to Vsphere eight for a lot of reasons, right? Not just or NBMECP or TCP. There's a lot of stuff that Vsphere eight has brought to the table here,
41:08
right? So with every version of V sphere, there are a lot of features that gets released, right? So if you can my recommendation from perspective and that's the party line that I'm towing here, please do go to V eight helps you a lot. All right. Add on to what I was asking around that.
41:23
So talked about snapshots with evolved. So you said cyber ransom critical thing out there as it stands right now. So with that being said, let's say I have that critical database, I want to do snapshots around this database. I need an hourly snapshot and I need to keep that for three or four days hours. How does that work with the evolves and technology as it stands right now that you guys
41:47
have? So you, you wanna take V balls, let's say if I have run V Balls in this book, when I need an hourly snapshot using balls, and I want to make that immutable snapshot using pure safe mode um using so safe mode is supported. Um And, and you're, you're saying in the context of safe mode for, for cyber, right? Um So it's supported.
42:10
There are some considerations we've got, you know, I, I hate to always point at the website, but the, the VM Ware platform page, we have a specific article on safe mode and um and best practices and some of the requirements. The real consideration about safe mode, snapshots and balls is um really one of object count considerations. Um The number of snapshots that you the length
42:32
of the snapshots, you know how long you're retaining them. Um The retention period is key, the count of the snapshots is key depending on what model flash array you're running. You know, there is an upper limit to the number of volumes and snapshots that the array can support. So you just have to be mindful of those things. Um Before you turn on safe mode,
42:52
before you move to, you gotta do a little bit of math and it's easy algebra and anybody from peer, even most account execs can help you do the math to make sure that you're counting, you know, a target number of evolves and a snapshot policy that will give you kind of peace of mind that you're not going to hit one of those upper limits if you hit one of those limits, like if you get close to the limits, we alert you,
43:14
we warn you that you never want to get in extremists and hit one of those upper limits. Uh We, we've had customers do that. It's not, it's not pleasant. We get, we get through it, but good to do the research earlier and don't ask me what the limits are for a specific model number. I I don't have it memorized.
43:31
So, but you can stump me if you do so. So uh we have uh an environment uh running on the VM ware virtualized uh with the pure storage flash, a broke uh fabric and running A B sphere seven plus. And lately we are getting running into issue where we find out that uh the rock rock
43:59
environment seems very sensible to any small blip in the environment. So we are struggling between each vendor because it seems to work in silos. Uh They give their recommendation and setting for best practice. But uh is there a way we can find uh a global best practice for that environment? And the answer is yes.
44:20
So please reach out to me after this so I can help you with that. So there is an Oracle rack deployment guide that's out there. There's an oracle best practices guide, there's an oracle B CD R guide, there's an oracle disaster recovery to the cloud. So, and I'll give you the one stop shop. Essentially, that's a blog link where all the
44:36
white papers I dumped it, right? But let me ask you a question. Have you opened up a case with VM Ware Gss? Please give me the ticket ID. Please give me the ticket ID because essentially all road leads to Rome. It will come back to me at some point of time, right? So what we have to do is look at, you know,
44:53
there are, there are some, there are best practices that needs to be implicated in every layer of stack, right? For example, you have a virtual machine with one PV controller and 25 em tickets. You know, and I've seen that I'm po I've seen that with custom environments. That's a bad way of how you can set a virtual machine up.
45:08
So there are there are ways that there are, there are actually best practices that one has to inculcate in the stack. For example, four PV controllers make sure you low balance of E MD K, right? A distributed port group for a public, a private half quality of service, Mt U Jumbo frames and all that stuff, right? So once we make sure those are set in place,
45:27
you should be able to be motion an oracle raster all day long. And we actually did that with 5.15 0.1 and oracle 11 gr two. So please do reach out to me and I'll help you. Thank you all for coming and thank you all to the panelists for coming. Uh If you have additional questions,
45:48
these fellows will be out in the hallway to answer your questions. Thank you.
