As we all know, EMC acquired VMware quite some time ago, and owns a large chunk of VMware to this day. The obvious reason that EMC values VMware is that the deployment of VMware vSphere in large accounts drives the need for and the purchase of large amounts of high end storage. This is very good for EMC, but it is a bit of a problem for VMware. The reason that it is a bit of a problem for VMware is that in many virtualization projects, acquiring the required storage accounts for over 50% of the cost of the entire project. Up until now there has not been a heck of a lot that you could do about this – you just had to live with it.
Now Microsoft is starting to make noise about the next version of Windows server, variously named Windows Server 2012 or Windows Server 8, and its included new version of Hyper-V, Hyper-V 3.0. Now one of the things that Microsoft loves to do to competing vendors is to commoditize their markets, attack their income statements, and thereby deprive those vendors of the oxygen (money) needed to fund the further development of competing products. This is precisely what Microsoft did to the Unix server market with the release of Windows NT 3.5 and the follow on release of the subsequent Windows Server products. Microsoft turned a very high end, very complicated, very specialized server market into a market with commodity servers running the Windows Server OS.
How Windows Server 8 and Hyper-V 3 Could Change the Game
By basically including Hyper-V in Windows Server, Microsoft has always tried to make the point that in their opinion, virtualization was a feature and not a separate product worth paying for. Clearly the enterprise virtualization market has not agreed with this point of view and has largely standardized on VMware vSphere – justifiably due to its superior feature set, performance, reliability, sociability, and broad ecosystem support. On the other had the SMB customers who are 100% Windows and who have only Windows Admin expertise in house, have seen Hyper-V as an easy addition to their capabilities.
The one common thread across both the enterprise virtualization market and the SMB/SME market is that the shared storage (often a SAN) required to make the really desirable features of vSphere and Hyper-V work (all of the features layered on top of vMotion, and Live Migration) work is expensive. It is expensive at a small scale for an SMB who just is supporting a couple of servers, and it is really expensive for large enterprises. So expensive that most large enterprises have a VP of Storage just to manage the expense.
With Windows Server 8 and Hyper-V 3, it appears that Microsoft is enabling some rather significant new price/performance points in the virtualization market. Here is what they are doing:
- There is a new storage protocol in this version of Windows, SMB 2.2. Now before you choke on your coffee or your sandwich and say that SMB has been awful in the past (you would be right), recognize that it appears that Microsoft has completely rewritten SMB to be something that really works well. As a result of SMB 2.2 it will be possible to use a simple file server as shared storage with one NIC in each host pointing to the file server. That means that highly desirable features like Live Migration will no longer need a cluster or a SAN to work. This will be true as long as both the file server and the hosts are all running Windows Server 8 and the hosts are all running Hyper-V 3.
- If you want your storage to be fault tolerant, Microsoft will be delivering a whole new type of fault tolerant storage called SOFS. This will allow one of the file servers to fail completely and have the VM’s experience no interruption of service (not like what occurs today in a cluster failover). Now you are not going to get all of those fancy storage features that come with the high end storage arrays that sit on your SAN with SOFS, but what you are going to get is more redundant storage for your dollar than with those existing solutions.
- For entreprises, SOFS is probably not going to replace the high end arrays (we need to wait and see what innovations the hardware vendors come up with to take advantage of SOFTS), but SOFS may well become another viable tier in the storage architecture. Being the best price/performance tier in the storage architecture – even if you lack all of the high end features is a great place to start, as you can reasonably expand functionality and eat into the higher end tiers over time.
So What is VMware Going To Do About Commodity Storage?
There are a couple of things that need to be stated about how VMware is going to react at the outset of this analysis. The senior people at VMware (Paul Maritz, Todd Nielsen and Richard McAniff) are all ex-Microsoft executives who successfully ran the “attack the income statement and commoditize a market” play on a variety of vendors while they were with Microsoft. Victims included Netscape, 3Com, Sybase, and the high end server businesses of HP, IBM, and Sun. So VMware understands this play perfectly is understands exactly how it is going to unfold.
The next thing is that there is absolutely nothing in this article that is going to be any kind of a surprise to the product and strategy planning people at VMware. VMware has known for a long time that Microsoft was going to do this and has, in fact, already released its first response. That response was the vSphere Storage Appliance, that allows you to use the storage local to up to three vSphere hosts to construct a virtual SAN (so not even a file server required to do vMotion).
The issue that VMware faces is that the Microsoft threat on the commodity storage front is far greater than the market for three host vSphere environments. No it does not extend all of they way up VMware’s largest customers – but if VMware continues to rely heavily upon expensive storage, and Microsoft is out revolutionizing the storage market by commoditizing it, then VMware is going to have a problem. This problem will occur if and when the cost of buying storage for a Hyper-V installation becomes significantly less than the cost of procuring that storage for a vSphere implementation – based upon VMware’s continued reliance upon expensive shared storage.
This is where the wedge gets created between VMware and EMC. EMC loves it that VMware is driving demand for expensive storage. VMware is going to have to start driving demand for cheap storage in order to remain cost competitive with Microsoft Windows Server 8 and Hyper-V 3. You can be certain that there are significant software development projects going on at VMware right now to address this issue and it is highly likely that we will hear about some of them at VMworld this year. Look for VMware to continue to put the features that distinguish expensive storage from cheap storage into the hypervisor and continue to decouple the requirements for expensive shared storage from the most important features in vSphere like vMotion, DRS, HA and SRM.
So What About VMware and EMC?
If the rationale for EMC buying and owning VMware was that VMware drives the demand for high end storage, and it becomes VMware’s product strategy (out of necessity) to drive demand for cheap storage instead then that blows up the entire financial rationale for EMC owning VMware. In fact it goes further than that. It creates a situation where being owned by EMC and EMC having any influence over the product strategy of VMware is actively harmful to VMware, and harmful to the long term ability of VMware to create value for its customers and its investors. So going way out on a limb here – here is prediction. Three years from now, EMC will no longer own a material part of VMware. VMware will be a fully independent company free to do exactly what Microsoft does, partner with vendors who want to commoditize things like storage in order to advance the price/performance point for virtualization. If that is the result of Microsoft Windows Server 8 and Hyper-V 3, then ironically Microsoft will have done VMware a huge favor.