I’ve been looking at public cloud offerings informally for a while and while I appreciate the case for a full cloud infrastructure, it will be quite some time before large enterprise datacenters can realistically retool everything for the cloud. Sure, small startups can very successfully have large parts if not all of their IT services in the cloud but there are still too many barriers for larger organizations with large investments in “in-house” IT resources.
My initial thoughts were that there could be an opportunity for hybrid clouds that can provide organizations with excess capacity on demand. This would be a great way to augment datacenters with cloud bursting opportunities and introduce organizations to cloud offerings with low risk. Of course, accomplish this, enterprises will need to build infrastructures that are compatible with public clouds.
The problem is that none of the (admittedly small sample) of private cloud infrastructures that I’ve looked at use the same APIs as their public cloud counterparts. Until now that is …
Here’s an interesting announcement from Eucalyptus and Terracotta that essentially provides the management tools and a private cloud infrastructure that uses the same APIs as the Amazon AWS. This is a good start and I hope and expect to see more offerings that make it easy to build hybrid cloud solutions.
I expect that MS will soon announce on-premises availability of the Azure platform.
I’ve been involved with Terminal Services architectures since 1995 with the classic Citrix WinFrame. I even spent some time at Microsoft as the Technical product Lead for Terminal Services in the 1990’s. I’ve always felt that TS was an overlooked technology in most North American enterprises. European enterprises seemed to more readily accept this desktop computing paradigm. It just makes so much sense from a centralized management of computing assets and less touch points.
Then along came Virtual Desktop Infrastructures (VDI) and everybody was excited about the opportunity to centralize desktop computing resources. My initial thoughts were that this is interesting and makes sense for some specific use cases but that TS made sense for more use cases in the average environment. TS provides higher user density on the same server hardware, there were less touch points (with VDI you still have to manage each virtual desktop), there is a rich set of tools to help scale and manage TS infrastructures, and of course the technology was mature and proven in thousands of installations around the world. In reality, TS and VDI are parts of the same spectrum of thin client computing technologies – TS is single multiuser operating system, while VDI is multiple single user operating systems. I was waiting to be vindicated – waiting for the masses that jumped on the VDI train to realize that VDI was just another flavour of TS and that TS made more sense for most users.
I was wrong. VDI is going to leave TS in the dust. Don’t misunderstand; I still believe that TS makes sense for many use cases and that there will continue to be a market for it. The sheer momentum behind virtualization initiatives has propelled VDI into the spot light for most organizations looking at re-architecting their desktop delivery strategies. That momentum might be enough to eclipse TS. After all, the best technology doesn’t always win out in this industry, for instance, <insert you favourite failed technology example here>.
I believe that VDI has one distinct advantage that TS can’t easily provide: It is more “cloud friendly”. Many enterprises currently have applications that can have their workloads dynamically moved within a virtualized pool of computing resources. This works well as long as computing resources are available to meet the peaks. This usually means that for a large portion of the time, the supply of compute power exceeds demand. Essentially, compute power is over provisioned at least some of the time.
However, using cloud services (public or private), can provide just in time computing resources that allow applications to burst into the cloud to meet peak workload requirements.
Think about having the ability to do this with desktop workloads. Sure you could do this with Terminal Services but the inherent TS advantage of higher user density makes the workloads less granular than the equivalent VDI workloads, and more apt to be over provisioned. Essentially, you could establish a policy that would let specified VDI users burst to the cloud as required. This would be perfect for occasional users, or low priority users.
What a great desktop delivery model – provision on demand. Couple this with a client side hypervisor, a streamed desktop, and local caching and you have a solution for the road warrior too.