Windstream Experiences the CAPEX and OPEX Savings of Diskless VDI; Interview with Atlantis Computing CEO Part III
Companies that implement a virtual desktop infrastructure (VDI) deployment may discover that it provides no greater performance and yet is more expensive than a standard PC deployment. That is, until they convert to a diskless VDI deployment.
That was the experience of Windstream, a large telecom firm that first implemented traditional VDI before switching to diskless VDI. The upgrade led to enhanced performance, lower CAPEX and OPEX costs, greater endpoint device options, and quicker integration of new employees and acquired companies. In this third and final part of my interview series with Atlantis Computing CEO Bernard Harguindeguy, we conclude our interview series by taking a close look at Windstream's experience with converting to diskless VDI.
Ben: Your diskless VDI solution has performed really well in tests with Cisco, and you've touched on the experiences of a couple of customers. Can you share any customer experiences in greater detail?
Bernard: Sure. We have a customer named Windstream that's deploying Atlantis ILIO. They acquired a company called PAETEC, a telecommunications company. They had some of the typical VDI drivers, cost savings, limited desktop refresh, centralizing all their data to eliminate data loss.
They wanted to provide for different types of end-point devices, such as iPads and a bring-your-own-PC program. And then they also wanted to be able to bring on board new employees and new companies that they acquired faster.
Now, they ran into some of the typical storage challenges with VDI. They had, when they looked at using their existing NetApp arrays for it, a very high cost of storage.
Also, they were getting limited IOPS. So they were spec'ing it at 10 or 20 IOPS per desktop, which is not going to give equal or better than PC performance, which is what their users were looking for and what they set as their kind of minimum bar. In addition to that, the cost was going to be far in excess of what a physical PC was when you added up all the infrastructure, the operational expenses.
Ben: How did they go from a VDI solution that was not working for them to yours?
Bernard: They started looking at different VDI architectures to try to figure out a better way to do it. First, they looked at local SSDs [solid-state drives] and PCI RAM [peripheral component interconnect random access memory] drives. But what they found was that the capacity was limited, especially in the high-density blades that they were looking at.
Also the costs of some of those cards were very high. So, it increased the cost per desktop.
The third problem was that they wanted to go with a blade server architecture. But their blades didn't support those types of drives.
So what they ended up going with was Citrix XenDesktop with Atlantis ILIO diskless VDI optimizing all the virtual desktop images, those running off of local Cisco UCS blade memory. And then thin clients, repurposed PCs with thin client software, iPads, and bring-your-own end-user PCs as the clients. They were able to see the benefits from a performance perspective and 11 second boot time.
Ben: That's a nice result.
Bernard: I'd like to go a little deeper so you can see the exact server specifications that were used by the Windstream PAETEC deployment and our recommended architecture with Cisco.
Ben: Sure, go ahead.
Bernard: In this case, on a single blade we used the VMware vSphere Hypervisor with 2 GBs of RAM allocated to that. We had 160 virtual desktops, which were Windows 7.
They each had 2 GBs of RAM. They shared 19 processors amongst themselves--19 of the 20. Plus, Atlantis ILIO as the virtual machine that ran alongside the virtual desktops.
Source: Atlantis Computing
The Atlantis ILIO software takes 6 GBs of RAM and one processor. And when it's installed, it takes the remaining memory. It creates a RAM disk, which is used to store the virtual desktop images.
Ben: That RAM disk, was it configurable?
Bernard: Yes. During the installation you can increase or decrease the size of the RAM disk. Typically it ends up being a function of how many virtual desktops you have on that server. Usually we recommend about one gig per desktop. That gives you plenty of "headroom" in terms of having enough space.
Ben: When the blades first boot up, they have to pull the images from somewhere?
Bernard: In the case of Windstream, they're using Citrix with provisioning server. So the images are provisioned on demand to new servers and re-provisioned after every session. So that automatically happens in the type of VDI architecture being stateless, or non-persistent.
Ben: So, you're really looking at non-persistent. So there's no reason for you to be caching back to disk at all. It can all stay in memory.
Bernard: And that means faster than local memory. It's an objective in the diskless VDI scenario to get the images small enough so that it can use local memory and not have to use shared storage. It's a little different approach technically. But it uses the same core technology.
It's important to note that the diskless VDI version of Atlantis ILIO is designed specifically for a non-persistent, or stateless, type of VDI deployment. So user data would be stored on the network somewhere as part of a profile management tool. So we're storing the virtual desktop images themselves on the Atlantis ILIO data store, which is what requires you to buy such a big storage array, typically.
Ben: What about a persistent type of VDI deployment?
Bernard: We support persistent with our traditional product. CBRE, the example I gave earlier [in part II], has an architecture that you can use with persistent. So you take shared storage, you use the normal Atlantis ILIO product, and then you can do that with persistent desktop images. A lot of customers have a mix of the two or start off with persistent and then migrate to non-persistent.
In Part I of this interview series, we discuss how Atlantis Computing optimizes storage for VDI deployments by actually eliminating the need for conventional disk storage.
In Part II of this interview series with Atlantis Computing's CEO Bernard Harguindeguy, we will take a deeper look at the workings of VDI storage optimization as well as reveal which architecture setup, according to testing with Cisco, delivers the best cost-performance results.
That was the experience of Windstream, a large telecom firm that first implemented traditional VDI before switching to diskless VDI. The upgrade led to enhanced performance, lower CAPEX and OPEX costs, greater endpoint device options, and quicker integration of new employees and acquired companies. In this third and final part of my interview series with Atlantis Computing CEO Bernard Harguindeguy, we conclude our interview series by taking a close look at Windstream's experience with converting to diskless VDI.
Ben: Your diskless VDI solution has performed really well in tests with Cisco, and you've touched on the experiences of a couple of customers. Can you share any customer experiences in greater detail?
Bernard: Sure. We have a customer named Windstream that's deploying Atlantis ILIO. They acquired a company called PAETEC, a telecommunications company. They had some of the typical VDI drivers, cost savings, limited desktop refresh, centralizing all their data to eliminate data loss.
They wanted to provide for different types of end-point devices, such as iPads and a bring-your-own-PC program. And then they also wanted to be able to bring on board new employees and new companies that they acquired faster.
Now, they ran into some of the typical storage challenges with VDI. They had, when they looked at using their existing NetApp arrays for it, a very high cost of storage.
Also, they were getting limited IOPS. So they were spec'ing it at 10 or 20 IOPS per desktop, which is not going to give equal or better than PC performance, which is what their users were looking for and what they set as their kind of minimum bar. In addition to that, the cost was going to be far in excess of what a physical PC was when you added up all the infrastructure, the operational expenses.
Ben: How did they go from a VDI solution that was not working for them to yours?
Bernard: They started looking at different VDI architectures to try to figure out a better way to do it. First, they looked at local SSDs [solid-state drives] and PCI RAM [peripheral component interconnect random access memory] drives. But what they found was that the capacity was limited, especially in the high-density blades that they were looking at.
Also the costs of some of those cards were very high. So, it increased the cost per desktop.
The third problem was that they wanted to go with a blade server architecture. But their blades didn't support those types of drives.
So what they ended up going with was Citrix XenDesktop with Atlantis ILIO diskless VDI optimizing all the virtual desktop images, those running off of local Cisco UCS blade memory. And then thin clients, repurposed PCs with thin client software, iPads, and bring-your-own end-user PCs as the clients. They were able to see the benefits from a performance perspective and 11 second boot time.
Ben: That's a nice result.
Bernard: I'd like to go a little deeper so you can see the exact server specifications that were used by the Windstream PAETEC deployment and our recommended architecture with Cisco.
Ben: Sure, go ahead.
Bernard: In this case, on a single blade we used the VMware vSphere Hypervisor with 2 GBs of RAM allocated to that. We had 160 virtual desktops, which were Windows 7.
They each had 2 GBs of RAM. They shared 19 processors amongst themselves--19 of the 20. Plus, Atlantis ILIO as the virtual machine that ran alongside the virtual desktops.
Source: Atlantis ComputingThe Atlantis ILIO software takes 6 GBs of RAM and one processor. And when it's installed, it takes the remaining memory. It creates a RAM disk, which is used to store the virtual desktop images.
Ben: That RAM disk, was it configurable?
Bernard: Yes. During the installation you can increase or decrease the size of the RAM disk. Typically it ends up being a function of how many virtual desktops you have on that server. Usually we recommend about one gig per desktop. That gives you plenty of "headroom" in terms of having enough space.
Ben: When the blades first boot up, they have to pull the images from somewhere?
Bernard: In the case of Windstream, they're using Citrix with provisioning server. So the images are provisioned on demand to new servers and re-provisioned after every session. So that automatically happens in the type of VDI architecture being stateless, or non-persistent.
Ben: So, you're really looking at non-persistent. So there's no reason for you to be caching back to disk at all. It can all stay in memory.
Bernard: And that means faster than local memory. It's an objective in the diskless VDI scenario to get the images small enough so that it can use local memory and not have to use shared storage. It's a little different approach technically. But it uses the same core technology.
It's important to note that the diskless VDI version of Atlantis ILIO is designed specifically for a non-persistent, or stateless, type of VDI deployment. So user data would be stored on the network somewhere as part of a profile management tool. So we're storing the virtual desktop images themselves on the Atlantis ILIO data store, which is what requires you to buy such a big storage array, typically.
Ben: What about a persistent type of VDI deployment?
Bernard: We support persistent with our traditional product. CBRE, the example I gave earlier [in part II], has an architecture that you can use with persistent. So you take shared storage, you use the normal Atlantis ILIO product, and then you can do that with persistent desktop images. A lot of customers have a mix of the two or start off with persistent and then migrate to non-persistent.
In Part I of this interview series, we discuss how Atlantis Computing optimizes storage for VDI deployments by actually eliminating the need for conventional disk storage.
In Part II of this interview series with Atlantis Computing's CEO Bernard Harguindeguy, we will take a deeper look at the workings of VDI storage optimization as well as reveal which architecture setup, according to testing with Cisco, delivers the best cost-performance results.
Buyers Guides
Newsletter Signup
Leave a comment