Home / Nerdio Academy / Nerdio / Nerdio Fundamentals: Nerdio Cost Estimator (Part 2 ITaaS)

Nerdio Fundamentals: Nerdio Cost Estimator (Part 2 ITaaS)

0 commentsJuly 22, 2019Videos

Joseph Landes
In this session we’re going to talk about an incredibly popular tool that Nerdio has to offer, the Nerdio cost estimator. We know that one of the biggest challenges MSP’s face in building a cloud practice in Microsoft Azure is deeply understanding the cost of purchasing Azure from their distributor, or in some cases directly from Microsoft. MSP’s are concerned that in moving to a consumption based selling model that they may accidentally quote their customers and incorrect, or unrealistic price for Azure services. The Nerdio cost estimator aims to solve this challenge by providing MSP’s with an easy to use tool that has considered all of the pricing challenges one may encounter when quoting Azure deals to their customers.

Joseph Landes
We have organized our cost estimator by the most common use cases an MSP would deploy Azure. In this session today we will walk you through the cost estimator looking at a more extensive use case, IT is a service where the customer is running a complete virtual desktop environment in the cloud. Enjoy the session.

Vadim Vladimirskiy
Lets jump into the coste estimator, you remember it is under the map under the billing menu and [inaudible 00:01:25] cost estimator is right here, and we’re going to be going through IT as a service. So IT as a service is for running a complete virtual desktop centric IT environment in the cloud. It’s designed to migrate an organizations full IT stack to the Microsoft cloud, run desktop servers applications, back-up and security in Azure, integrate with email, collaboration, and file storage and office 365.

Vadim Vladimirskiy
First thing you’ll notice is as we select the IT as a service use case, then our selections change and the additional sections and additional options. So lets go ahead and go through an example. Once we select IT as a service then the next section is going to ask us about the desktop information, because this is a use case that is desktop centric, there is going to be a significant portion of the environment that is going to be focused on users desktops, so lets start with an example.

Vadim Vladimirskiy
Lets say we’re going to have a hundred person, deployment or a hundred desktop deployment to be precise, not necessarily a hundred users. There’s nothing in here that’s user centric, it’s really all about desktop user centric. Okay, so this is going to be our total number of desktops, or lets go a little bit bigger, lets go with one fifty as an example. All right, then the next thing is, out of those a hundred and fifty, how many of those users will need dedicated VBI desktops? And if you mouse over the tool tip on the right it says executives and managers often use dedicated desktops, these desktops received dedicated compute resources and users are fully isolated from each other.

Vadim Vladimirskiy
If we have a hundred and fifty person organization, and we may have twenty of the people using VBI dedicated machines, that means the remaining users, which is basically one-fifty minus twenty, are RDS users and when we talk about RDS users we mean that these users are going to be sharing session hosts, set up session hosts, RDS collections, et cetera. They’re not going to get their own dedicated resources.

Vadim Vladimirskiy
Then the question over here refers to graphically intensive users with, or graphically intensive desktop needs, so this would be for GPU instances. If we type in the number here it’s going to take that number of users, so lets say were going to place five users there. What you’ll notice is now we have a Hundred and fifty total, twenty VBI, five GPU RDS and a hundred twenty-five standard RDS.

Vadim Vladimirskiy
The next selection is the type of plan that’s going to be used for Nerdio. So there’s two plans, there’s enterprise and professions. The reason I selected such a big environment is because it’s really not a fit for a small deployment and that’s what professional is all about, so this [inaudible 00:04:29] organizations with more than fifteen users, could be up to twenty-five in some cases, enterprise edition requires a minimum of seven cores in your Azure subscription typically used for large organizations that want multi-factor authentication, DR viral management, RDS collections and hybrid AD.

Vadim Vladimirskiy
We’ll go with enterprise and then we’ll do one more example later with professional as well. Okay. So that’s what we need for the desktop side of things, then we go into server side of things and there’s really two questions here, one is- How much total storage is going to be needed for files and databases? So this will, basically, be the size of the data disk attached to the file server and there is some pre-configured selections here. So anywhere from a hundred twenty-eight to eight terabytes. Lets say we’re going to go with the two terabyte disk, and then the next question is going to be- will there be any application or databases that required dedicated servers?

Vadim Vladimirskiy
What this means is that, in addition to the basic environment in addition to domain control and file server, RD gateway, RD session host, VDI desktops, et cetera. In addition to all of those, will there be any applications or sequel servers, or any of those additional VMs? And if we select yes, you’ll notice a familiar layout where you can go in and add various servers. Lets go ahead and just add a couple servers here. We’re going to do sequel 01, it’s going to be a D4B3. It’s going to have a P-10 OS disk and a P20 data disk and then we’re also going to have an application server, that will be D2B3 with an E10 for OS and no data disk at all. Okay.

Vadim Vladimirskiy
So now we got two servers in addition to whatever the system is going to quote out for everything else. Okay. The next sections is about licensing. So the two questions here are about office 365 E3, so you should all be aware of the fact that in order to license office installed on the desktop you need version of pro-plus and office pro-plus is included as a subscription, either in a stand-alone skew of office pro-plus, or in an E3 or E5 skew. It’s not included with E1 and it’s not included with Microsoft business or office business premium, right, those don’t include pro-plus and pro-plus is the only one that’s supported with the RDS role enabled because it needs shared computer activation.

Vadim Vladimirskiy
So if you want to include, if you are the partner and you’re doing this analysis for a hundred and fifty users, you may well include a hundred and fifty licencees of E3, in this particular estimate. So lets go ahead and select a hundred and fifty users and then the next question is about RDS licenses, so because we’re going to be deploying RDS, whether or not users are going to be getting dedicated VBI desktops or using shared session hosts, every user needs an RDS subscriber licencee, it’s per user licencee, and there are three ways to provide that licencee.

Vadim Vladimirskiy
Number one- if the customer all ready owns and RDS licencee, could be they purchased it for an existing RDS environment that’s on prime, or in the dedicated data center, in that case if they have software insurance then they are able to bring their own licencee, which is when you would say no, in this case. The other scenario is if the partner is a [inaudible 00:08:28] partner, meaning that they can rent Microsoft licencees out, not through the CSB program necessarily but being a [inaudible 00:08:36] partner then selecting yes here will add the cost of that licencee under [inaudible 00:08:42] into that analysis.

Vadim Vladimirskiy
There is a third way to now licencee RDS which is through CSB software subscriptions. That is something that’s going to be added into the cost estimator in the very near future, but for now this only accommodates either software insurance, meaning the customer brings their RDS license, or a [inaudible 00:09:02] licencee which is provided by the partner. Lets go ahead and say yes to that. Okay then, we’re down then into the other features section. You’ll see a lot of these things are going to be familiar but some of them will not, so we’ll go through each one, so we have this hybrid usage, so lets see what type of effect that has going from no, to yes. We’re going to go from forty-one forty-two down to twenty-five ten.

Vadim Vladimirskiy
So a significant sixteen hundred dollar or so savings, but we’re going to need four hundred and twenty-four cores of windows server and four hundred and twenty-four cores divided by eight, so we need fifty-three eight core packs, and each eight core pack for three years is about fourteen dollars a month. So we’re looking at seven forty-two per month, which means that it does make sense to go with hybrid usage, pre-pay for it up front because you save about fifteen hundred dollars and then you have to pay extra, about, seven hundred fifty, so you’re saving seven hundred fifty dollars or so by going from no, to yes and bringing your own server OS licencee.

Vadim Vladimirskiy
The next question is a gap regarding reserved instances, lets see what impact that has. So if we go from a three year reserved in twenty-five ten to pay-as-you-go, so we are increasing that by about fourteen hundred and twenty dollars or so, so that’s also significant savings. The next thing is in region back-up, i’ll remind you assumptions behind in region backup is that every VM is backed up. Every VM has fifty percent utilization of its disk and the data churned rate is two percent per day, and there is a thirty day retention. So from twenty-nine to down to twenty five, so about four hundred and forty, or four hundred fifty dollars is going to be the cost to back-up that environment. This entire environment with a thirty day retention and daily backups.

Vadim Vladimirskiy
Other region DR, say yes to that, you’ll that’s going to increase the price by slightly over a thousand dollars, that’s in addition to the backup into another Azure region with Azure site recovery, lets leave it off for now. Site-to-side VPN, for a large environment that’s typically needed, so we’re going to select site-to-site VPN that adds a small additional amount for a VPN gateway, a basic VPN gateway, which is about twenty-six, twenty seven dollars list price per month. Then the question is on hybrid AD. So hybrid AD, again you’ll recall is the ability to extend an existing active directory into Azure, by spinning up another domain controller and giving that visibility of that active directory, and being able to assign desktops and manage those user objects, that’s a common feature for large deployments like this, so lets say yes to that.

Vadim Vladimirskiy
As you’ll see me selecting yes, VPN now became a required field, and you’ll also know this I now have a server I can not remove from this section which is that external AVVC01, lets make is A2, a B2 MS two core, eight gigabyte of RAM VM, with an E10 OS disk. Okay, keep on going down. Okay, we have this option for desktop order scale, and you’ll notice that it’s grayed out, now why is desktop order scale grayed out? You’ll recall that we, when we discussed order scaling, there are two types of scaling. There is up and down, which means that the VM gets sized, resized during business hours sized up, and then off-hours sized down.

Vadim Vladimirskiy
And then there’s also a way to run RDS collections, which is a scale out and scale back in. Now with sizing a VM up and down, it Doesn’t make sense to do that when you are pre-paying for reservations. So if you’re prepaying for a three year, or one year reservation, then there is no sense in actually turning that VM off, or making it smaller because you’re already going to be pre-paying for that capacity and there’s no cost benefit by turning it on and off. However, if you are paying as you go, as opposed to reservation, you’ll see that this option will get highlighted and then as you make the changes, you can see what impact it has on cost.

Vadim Vladimirskiy
Since we’re going to be doing three years, lets leave this as no, because that’s our only option. Finally the less selection in this section is the on RAM regions. You’ll recall the on RAM regions is a way of spinning up a small Azure deployment with an RD gateway, which then gives you the ability to place additional desktops or server VMs in that region or simply route the user through that region to the primary desktop deployment. If we say yes here, we’re going to get a selection down there asking us how many on RAM regions there will be. We can say yes and there are going to be two of them, and what this accounts for is additional bandwidth, and it also accounts for the additional site to site VPN tunnels and VPN gateways that are going to be necessary to accommodate on RAM regions, so I’m going to say no to this for now.

Vadim Vladimirskiy
All right, one thing that I missed is this RDS collections options. Why am I… There we go. So we have this RDS collections that’s checked by default, and then most scenarios it will actually be used by an environment that has a hundred and fifty desktops in it, so lets keep that as on. If we turn it off you’ll see some of the UI will change, but lets turn it on and then i’ll explain how that works. So it says we noticed you like to take advantage of RDS collections to handle desktop workload, we will maintain the minimum of one, which is configurable, and no more than twenty-five, which is calculated. And it tells you it’s calculated based on the hosts size, and the number of users and users to CPU core ratio, and i’ll show you what that means.

Vadim Vladimirskiy
Then you can configure what kind of host you want to use, lets say if we are by default it’s a single core host, which is unrealistic. Generally that’s not going to be the case, lets say were going to go with an eight core host. So you can see if I’m going with an eight core then our maximum number of servers has gone down from twenty-five to four, and what happened with our price? Lets see we’re currently at twenty-eight nineteen, and here we’re at thirty-thirty, so it’s a little bit more cost efficient, not dramatically so, but if we go with four core VMs it looks like that actually makes a pretty significant difference. So, what’s the difference? Only few dollars.

Vadim Vladimirskiy
As you would expect, because the cost of Azure compute is pretty directly proportional to the number of cores, and we’re making these calculations based on how many users, how many users per core, and how many cores each VM has, so as you switch from one VM to another and all you’re modifying is, really, the number of cores this number changes accordingly and then the cost doesn’t fluctuate much, right? It fluctuates a little bit but not significantly.

Vadim Vladimirskiy
Okay, and i’ll continue talking about this because there are a few more details about RDS collections, but let me jump to this set of sections. This is asking a question of how many hours out of a week will this collection be maxed out? Meaning if we could have a minimum of one and a maximum of seven, and we can even say, “Hey, we were never going to have a minimum of one, We’re going to have anywhere between three and seven hosts in the collection, we’re going to have nine- lets say eight to six, which is ten hours, five days a week. Which is fifty hours in a week, is going to be maxed out, meaning it’s going to be at seven. Because that’s the business hours that we’ve decided we’re going to put in here, and then the remaining number of hours, so there’s still a hundred and sixty-eight in the week, so a hundred and eighteen hours are going to be the minimum size.”

Vadim Vladimirskiy
So for purposes of the calculation this is going to say- all right the cost that we’re going to calculate is going to be three hosts of this size running full time for fifty hours and then, I’m sorry, seven hosts of this size running full-time for fifty hours, and three hosts of this size running for a hundred and eighteen hours. Okay, that’s how the calculation works, and you’ll see on the next screen the assumption that is being made is that the minimum number of hosts that we program in here, will get reserved instances reserved just for them, and then the remaining hosts will be used without reservations. Meaning they will turn on and turn off as needed, so in this case we’ll have three reservations and four hosts in addition to those three. So we’ll have a total of seven hosts, but only three of them will have a reservation associated with them.

Vadim Vladimirskiy
Okay, then we go into the cost assumptions, we have bandwidth, so in the previous example we looked at, which was line and business servers, we could not do bandwidth in terms of desktop users because there was no users involved in that type of use case, here, since we are primarily focusing on desktop users, the type of bandwidth you need for a typical RDP connection is between a dollar and three dollars and it’s closer to the dollar so we’ll put it at one fifty, could be modified if needed. The storage operations, as we discussed last time, is a per gigabyte additional charge for the operations only applies for standard storage. Premium there isn’t any charge on those, so you just put in this average number and it will recalculate everything for you.

Vadim Vladimirskiy
Now this number is really important, this is what is really driving the calculation of how many hosts you will need, so lets use this as an example. This is configured and says that there will be an average of five users that can be handled by a single core on a standard RDS session host. So that means that if we have a hundred and fifty users we’ll need a total of thirty cores and since we’ll need a total of roughly thirty cores and each one of our VMs is four cores, actually I take that back, we don’t need a total of thirty cores because we are a hundred and twenty five users, because twenty are going to be dedicated VDI and five are going to have GPU, so they’re going to be on a different session host, so one twenty five over five is twenty five, so we need twenty five cores and twenty five cores divided by four, and rounded up, comes out to seven, right?

Vadim Vladimirskiy
So twenty five by four is obviously not an integer, divisible number, so we would round it up to seven, so that means that we are going to allocate up to twenty-eight cores for this collection that is going to support a hundred and twenty five RDS users. Now, where is this number coming from? This number is coming from some published research and our experience on what a typical density of a user per core happens to be. we’ve seen it ranges between, lets say, three and six, so if we want to throw a little more horse power into this we would say- lets say, four, and if you say four, right? What you see now is that, for the same size of a session host, we now have more of these session hosts that are going to be estimated in this analysis, so if we go three it’s obviously going to increase that even more, so lets just go with four as an example.

Vadim Vladimirskiy
Okay. So we’ll go with four session hosts, I’m sorry, four users per CPU core. Okay, this next selection applies to RDS session hosts with GPU, so recall we selected five users, are going to have GPUs. So in this field, if we say that two and a half users can be up within a single core, that means we need a minimum of two cores, and because the smallest GPU enabled RDS session host is six, this makes no difference. If I change this from two and a half to one, you’ll notice that my price doesn’t change at all, I change it to six my price won’t change but as soon as I change it to seven or more, lets see how much- no I’m sorry I need, actually no, it needs to be- sorry, it needs to be less than one in order for me to have more than six cores. Lets just say, one, as an example, that means every user will get a minimum of a single core on a session host, the next selection is the size of a VDI desktop.

Vadim Vladimirskiy
We’ve selected up here twenty users with VDI desktops, and here we get the opportunity to actually select what kind desktops will they have? If this is a standard user then maybe a single core VM is enough, but generally I like dual core VMs, and the minimum- so I would go with a B2MS which is two cores and eight gigs of RAM, lets leave that selected, and you see the price has gone up a bit because we’re now going from a D1V2 to a V2MS. Okay, then we have the familiar fields of the discounts, if someone is going through distribution they’ll usually see eight or nine percent, the reserved instances discount may be something like two percent, their office 365 discount may be something like fifteen percent, then you select the region where it’s going to be. So lets say it’s going to be in East US too, that made actually quite significant difference, what is it? Twenty-seven forty-nine, versus, that’s twenty-six but our, the GPU is not available in the region whether we have south central, south central, yep. There we go. South central is thirty seventy-eight, and it looks like East US happens to be… Wow, that’s really three hundred dollar difference which is obviously significant.

Vadim Vladimirskiy
And then the currency. All right. This is a sample configuration, again a hundred fifty users, twenty VDI, five GPU, two terabytes of shared storage, a sequel server and application server with these parameters including office 365 pricing and RDS licencees, using your own, bringing your own windows server licencee. Using three year reservations with a backup without DR, site-to-site VPN and hybrid AD, using RDS collections without desktop order scale, because the reservations make that unnecessary. No on RAM [inaudible 00:24:52] between three and eight session hosts of four cores at sixteen gigs of Ram each, and fifty hours of the time that collection of the, fifty hours of the week the collection is maxed out, a hundred and eighteen hours it’s at the minimum size. A hundred, no, dollar fifty per user, and bandwidth three pennies, three cents per user for standard storage for operations. Budgeting an average of four users per core on an RDS session host, one user per core on a GPU enabled session host, two cores, eight gigs of RAM, four individual VDI desktops, nine percent Azure discount, two percent reservation discount, fifteen percent office 365 discount running in the East US and priced in the US dollar currency.

Vadim Vladimirskiy
Okay. So all of that, we click on view costs we get this nice little break down, so now it tells us that the top level we’re looking at fifty-three dollars per user, per month. Which is the total of seventy-eight ninety-four, including licensing for Nerdio, including RDS, including office 365, and Azure. The thing it doesn’t include is the price for this, this is coming in the future version of the cost estimator but for now they will have to go off quote here and figure out what the cost of these cores will be, as we looked at, what was it? Two ninety-six over eight times, about, fourteen or fifteen dollars will be the cost per month for those licencees.

Vadim Vladimirskiy
Recap of the assumptions we made, two terabytes a hundred fifty-three, three year reservation no DR et cetera, and then here is the actual architecture of the environment. So we’ll have a domain controller and the file server ADFS proxy, hybrid AD controller, sequel server APSO 1 server, the ones we configured, an RD gateway, RD connection broker, a signal RDS sage host that’s going to have a GPU. It’s a six core host, so it’s plenty to accommodate the five users we each need one core. Twenty VDI desktops of two cores and eight gigs each, and then RDS session host between three and eight, and you note here it will tell you how many reservations you actually need, so all of these are reserved, with a exception of, in this RDS host analysis you have three hosts that are reserved, and five that are not reserved because they go in and out, and there’s no reason to pre-pay for them.

Vadim Vladimirskiy
Going down here it breaks this thing down into sub-components, so you can see about a thousand dollars is the desktops, which is really the cost of VDI plus the GPU users, RDS collections is additional four hundred and forty-four dollars, you can see fifty hours per week maxed out at eight. A hundred and eighteen hours per week it’s going to be at the minimum size of three, you can see the discounts. The site-to-site VPN cost, the regions’ server compute cost. Bandwidth and IPS et cetera. And then the total average cost per month is twenty-seven forty-nine, but because we’re using reservations forty-four thousand three hundred and one dollars will be paid up front, in the reservations. Then fifteen hundred and sixty-two dollars will be the cost per month for the rest of the time.

Vadim Vladimirskiy
Okay. Now this- Lets just roll it all up into this one number, which is fifty-three dollars per user. And we used pretty aggressive settings in terms of saving. So lets say this is, what we call, the best case scenario for a partner, when they go to sell the solution to a customer this is what it’s going to cost them, their cost, right? Best case scenario, now lets look at some of the assumptions that we made, and lets put them more towards list prices. If a customer were to go online and to Azure and buy this thing directly, how much would that cost? So what we need to do is change this to a no, change this to a no, and then remove all of these discounts. So basically set this to zero, set this to zero, this got grayed out because we’re not using reservations anymore, and what you’ll see is list price for the environment. Or what we like to call the worst case scenario for the partner is going to be seventy-nine dollars per user, you can see Azure is obviously much more significant component in that case.

Vadim Vladimirskiy
So if, when they go and try to figure out their proposal to the customer and price it out, they should be using this number as their starting point and then once this environment, once the customer is using the environment they can go and then optimize it by buying this, buying this, and applying their Azure discount to an office 365` discount, to a thing, right? So it’s very significant difference, it’s a twenty-six dollar per user per month difference, times a hundred and fifty users between the best case and the worst case scenario.

Vadim Vladimirskiy
And hopefully you can see with the Nerdio for Azure cost estimator, this type of what if analysis becomes really easy, so you can go back and you can see what the impact of making little changes is, is it the big lever or small lever? And something like this to create it from scratch using the Azure calculator or some other tool would be, a lot more complicated and a lot more work.

Videos in the series