[MUSIC] >> The first thing Ialways recommend with like if you're doing VMs on Azure is, figure out if you really need to, what's the size? Whatdo you actually need? I think we are offering almost 300 differentsizes and types of VMs.
Those are optimized fordifferent scenarios.
So some of them are optimizedfor small workload, some of them for CPUintensive workloads, we have some memoryintensive workloads, we have some with graphic adapters, some with ultra-fast storage.
So you need to figure out whatis actually the size I need.
What I realize is a lot of peoplealso who migrate VMs to Azure, they just take the size theyused on-prem they have before.
Turns out that a ton of theseVMs were just oversized.
You have probably seen that before.
I have a domain controlleror a Web server here.
I'll just give it like 16 gigs of memory because we have it anyway.
In Azure, you pay for this.
Or in every Cloud, you start paying for what you'reactually using.
So if you can create a smaller size of that VM, that's really helpful.
So people may ask now, Thomas, how can I actually find outthat my VM is too large, my on-prem VM, for example, before I move it to Azure? Turns out that Azure Migrate, which is a product whichwe have or a service, if you will, which helpsyou migrating VMs, but also databases andother workloads as well can go out and measure your on-prem environmentand does an assessment of it.
It can go and do performanceanalytics over time.
So it can tell you, hey, that VM has four cores.
But actually, youonly need two cores.
It's not that the VM everneeds that much resources.
So now I see you are, I thinkon the Azure Migrate page, that's exactly thegreat thing to use, even though if you're not migrating.
Again, this is free.
You can add some thirdparty and they will cost some additional money because they have additional capabilities.
But you can do a free assessment of yourenvironment without moving anything, just to see, hey, how I'm actually sized on-prem.
Then you can take that knowledge.
I basically start migrating orcreating the VMs as you want it.
That's only when you have on-prem VMs which youwant to move to Cloud.
I'm sorry that I have jumpedthere, but I have to.
I mentioned that as well.
If you already have VMs in Azure, we have something fantasticcalled the Azure Advisor, which can go out and does thesame thing and tells you, hey, this VM might be, not size the right way.
There's a lot of differentrecommendations, but it can also dorecommendations on sizes.
So as you said, pickingthe right size is the first thing which you should do.
You should not just select something because youthink that matches.
You should reallyunderstand what to pick because that will give you a much more cost effective solution.
>> Yeah, exactly.
It's funny is that I have acouple of people and one of my sessions and I was mentioning some of the free stuffand they were like, oh, no.
You can't host an app servicewithout paying money.
Actually, well, that's free as well.
What's happening there wasthat because we want you to have a high availability and we expect a lot of people usingapp service or other ones, it may be a company website, it may be even more, and so we guide you into a place to where custom domains and some of the other stuff is available for you.
But there's also the free skew that I hardly ever hearanybody really talking about.
I use the free skew for alllittle personal, little projects.
It is a shared instance, but it does give you a tonof information about how to get up and get started andstart building with Azure.
>> Absolutely.
I cannotstress this enough.
A lot of people don't knowabout this free offer.
Really, you should have alook at that to get started.
You can run a coupleof different things.
If you have a smallproject, you can do that.
I think coming back inthis Azure Advisor thing, what the question also wasabout is that shut down and turning off andrunning it in a way.
The next thing really, to pick that sizes, it's also about pickingthe right model.
So of course, theCloud is great if you have that agility and elastic, it can basically go andsize it the different ways and you can shut itdown and you save money.
So that's alsorecommendation obviously, as mentioned in the question, that you shut down your VM becausethen you only pay for storage.
However, that wasalso in the question, is that a lot of workloadsdo not really work that way.
I mean, I cannot just shut off basic services whichneed to run 24/7.
But for these VMs, we have something which for strangely not a lotof people know about, and it's called Reserve Instances.
So if you know thatyou have a VM which basically runs 24/7 for ayear or even three years, you can do basically apre-commit for that.
Or you can basically commit to have that VM running fora year and you pay, and you say okay, and you get ahighly reduced costs for that.
So yeah, you pulled up thewebsite, here you can see that.
The great thing is now I think we did that change a couple of months back is that before youhad to pay that upfront, so you have like, okay, I know I'm going to run this VM for three years.
So you get a highly discounted VM, but you needed to pay that upfront.
I said now we change it.
Now you can pay that.
You get still get the discount, but you can pay that cost over time, like in a monthly fee.
Basically not to justpay everything up front, which makes a lot ofcustomers very happy.
I think that's is where I saycompanies who run the lot of VMs, either you run them like 24/7, or you run them for a momentand then shut them down.
So those are two main differences depending on the workloadthat you run in that VM.
>> Yeah, exactly.
Gregor made a great comment aboutit in monthly as well now too, just pulling that out monthly.
Some Cloud providersallow you to write SaaS VM simply basedon RAM cores, etc.
You can set it to four cores andscale around without limits.
Azure uses preset fixed RAM combos.
Why wouldn't Azure allowcustom core RAM ratios? I've never actually tried to do anything like that before.
Have you? >> I remember when I didrun for some companies, when I rerun theirvirtual machines on-prem, we could really go out and fine tune everything into thelast configuration.
I think the problemwe have with that, and I believe that, again, with those, I thinkit's currently 300, it could be even moredifferent sizes.
We really try to make it as good as possible for you topick the right size.
There should be a size forevery of your workflow.
Yes, maybe sometimes we don'thave exactly what you wanted.
But this is also a littlebit to do with how we do capacity managementin the back-end.
Imagine if every one wouldjust deploy VMs with one course and twoterabytes of memory, we will not be preparedfor, in that case, maybe not have the righthardware exercising.
So what we did is actually wehave different hardware types and then we mapped different VM sizes to it so that wealways have enough capacity for all these different cases.
I really hope, it wouldbe interesting to know what VM size and whatscenario you're missing.
That would be very interesting to know so we can have a look at that.
>> Yeah.
Also there is 290.
>> Okay.
>> Yeah.
There is 290.
One of the other thingsI was going to mention here was that I pickthe Windows 10 Pro.
We have a bunch ofother Windows machines are using in just a virtual machine.
As like you're lookingthrough this, well, this is filtered to thecurrent standard size, but there is quite a bitof things to pick from.
But then I noticed that down here at the bottom was thispricing calculator.
So there is a crossingcalculator that comes with that.
Yeah, it would be interesting to know what exact scenario isn'tavailable in this option.
>> By the way, this is interesting, what do you just have opened there.
You see the VM sizeswhich start with a B, this is also probably somethingto think about what we have here.
I always think a little bitabout two scenarios there.
One is the burstable VM.
So you can just see here, which basically allowsyou to like say, hey, your VMs usually don't reallyutilize the CPU that much, but at one point, they need it.
So we basically giveyou a cheaper VM, because you don't utilizingthe CPU that much.
But then you can burst it outand when your VM needs it, they can get moreperformance in that time.
So you can also lookat solutions like this where we havedifferent machine types.
But I also should mention is, the new offering we havewith Sport Virtual Machines.
So think about where you have a really large task which youneed to do some calculations on, and you need a lotof different VMs and resources basicallyto calculate this.
You don't care if a VM diesat one point and you just add more and more because thecalculation just goes forever.
So Spot VMs are a reallygreat solution for that.
So we give you again, highlydiscounted VMs, which you can use.
But when someone elseneeds the capacity, we'll just shut them down andyou will get them back when there is more free capacity.
So Spot VMs are really good also for certain scenarios, not for everyone.
Don't use it for your report.
But critical workloadswhich need to run 24/7.
But if you need to go in and have some calculationchops you need to do, then Spots VMs can bea great thing as well.
>> Yes.
So this is very interesting.
I have not worked withthe sourcing this jet.
It looks like this.
Is this a preview type of thing or has this beenout for a long while? >> It's not a while.
I'm not sure if you'recurrently a close look.
I think [inaudible] , but it has been the last coupleof weeks or maybe months.
But it's not that the old yet.
I think I tweeted about it, but I can't remember when it was.
>> So with this Spot VMs, everybody should thinkmore along the lines of it's a pay as yougo type of thing, wherever you decidethat you want to start, say you have a spend and like $200, it would run until you hit that threshold and thenit would just shut down.
Is that like the primary benefit? >> Yeah, exactly.
Soyou basically get it.
But if someone elseneeds the resources, we just can take it away and you get it again when we don'trun that much anymore.
So that's where you get it, and then you have obviouslya way better price for that.
But again, your workloadneeds to support it, otherwise, people aregoing to be very upset.
>> Got you.
[MUSIC].
