One of the time honored strategies in the computer industry is to package your competitor out of business. Vendor A thinks that they have a robust standalone product business, and Vendor B makes Vendor Aâ€™s product into a feature of a larger product and Vendor A goes away. There are many examples of this. Microsoft started this game by combining Word, Excel and PowerPoint into a Suite, putting the then incumbent leaders (Lotus, WordPerfect and Harvard Graphics) under severe pressure. Years later Microsoft did this same thing to Novel by packing Network Operating System functionality into Windows, and then a few years after that did it again to Netscape by packaging the browser into Windows.
Now both Microsoft and VMware are trying to do this to each other, but in more sophisticated and insidious ways than has ever been attempted before. The reason for all of the sophistication is that this is quite literally a war of extinction where the loser may end up going away. Letâ€™s take a look at the two strategies.Â
VMware and the Data Center OS
With the Data Center OS strategy, it is VMwareâ€™s short term intention to own the hardware interfaces to the physical resources in the data center. This means that the drivers for CPU, memory, networks, and storage are all under VMwareâ€™s control and ship as a part of VMwareâ€™s products. This alone is an ambitious undertaking worthy of an entire company for a long period of time due to the rapid evolution of hardware in the data center market, and the multitude of vendors that contribute innovation and products to this market. If VMware succeeds at nothing more than this, then VMware will have inserted itself into the stack between the hardware vendors and the traditional OS vendors (Microsoft and Red Hat for the most part). Owning this layer of the stack would ensure that VMware was at the least around for a very long time, as once drivers for a variety of complex hardware are working in a demanding data center environment they are not easily replaced.
But VMwareâ€™s ambitions go much further than this, and this is where the existential threat to Microsoft arises. If you think about the core functions of an operating system, they are basically to provide an interface to hardware, to schedule resources, and to provide services to applications. VMware today already does the first two of these quite well â€“ leaving for the most part the providing of services to the OS vendors whose products run in VMware guests. However, VMware is already taking steps to fill in that role of an OS, not necessarily by providing their own layer of applications APIâ€™s, but by taking steps to devalue the applications APIâ€™s from Microsoft. Here are some examples of these steps:
- VirtualAppliances.net has a virtual appliance which contains an Apache Web Server. This appliance is based upon Gentoo Linux, so there is an OS involved, but it has been reduced to its minimum possible size so as to contain just what is needed to support the Apache Web Server. There is definitely not a need for a Windows server OS, nor a server OS license.
- BEA, the vendor of the market leading J2EE applications server Weblogic has a product called BEA Weblogic Server 9.2 Virtual Edition. This is a virtual appliance that contains the Weblogic applications server and â€śBEA LiquidVM, a virtualization-enabled version of BEA JRockit JVM that can run on a hypervisor without a standard OS, allowing Java applications to run directly on the virtualized hardware. LiquidVM takes the places of the traditional operating system environment, greatly reducing the memory footprint of the virtual machine and improving hardware resource utilization. This also drives down maintenance and support concerns and costs, increasing agility and flexibilityâ€ť. Learn more about this virtual appliance inÂ the VMware Virtual Appliance Marketplace.
- JumpBox has released a virtual appliance that provides a full execution environment for Ruby on Rails applications. Ruby on Rails is a next generation Web 2.0 applications development tools that allows developers to be highly productive in the creation of next generation Web 2.0 applications. The JumpBox for Ruby on Rails appliance gives these developers an easy way to deploy their applications in a guest without having to worry about configuring a Ruby on Rails run time environment and its underlying OS. You can download this appliance also from the VMware Virtual Appliance Marketplace.
- If you need a back end database for your three tier application then VirtualAppliances.net has both PostgresSQL and MYSQL virtual appliances available for download on the VMware Virtual Appliance Marketplace.
The bottom line to the above set of products is that you can build a rich Web 2.0 application in Ruby on Rails, and run it on a web server and a database server and not require an OS license or a support agreement from either Microsoft or Red Hat. I think VMware will continue to push the envelope here and over time continue to use the open source community to allow an ever richer set of applications functionality to run directly in a Guest. It is highly like that the long term future of the vApp functionality that is in VSphere 4 will come into play here as well. So it is clearly the strategy of VMware to devalue the core features of the operating systems from vendors like Microsoft and Red Hat by replacing 2/3rdâ€™s of what they do with feature in VSphere and by using the open source community to devalue the applications server layer of these operating systems.
The Microsoft Counter-Strike
Microsoft has a two pronged approach to in turn make VMware irrelevant. The first is to simply build virtualization into the OS so that Windows Server contains the virtualization features that VMware claims are unique and essential. While Microsoft has a decent set of first generation virtualization functionality built into Windows Server today, and it expanding that set of functionality with Windows Server R2 (due out by the end of 2009), most people look at the Microsoft offering and see a number of holes relative to what VMware is offering. Those holes include memory over-commit, vStorage functionality, vNetwork functionality, the security functionality of vmSafe and vShield Zone, as well as the general performance and robustness of VMware. So while one could project that Microsoft would at some point in time catch up with VMware and make â€śfreeâ€ť into â€śgood enoughâ€ť that point in time is right now still off in the future a bit.
But there is a second arrow in Microsoftâ€™s quiver. What if it was no longer necessary to virtualize the server OS at all? What if you could containerize the server based applications than ran in Windows Guests today, along with all of their version specific supporting Windows middleware. What if you could have a Data Center Edition of Windows running on the same kind of massive hardware that VMware runs on today, and that one instance of the Data Center Edition of Windows could support 50 or 60 different applications each requiring its own version of IIS, or SQL Server, or networking communications support (all of these things being part of the containers). Â Microsoft recently demonstrated this precise functionality in an Application Virtualization Demo at TechEd and then also at the Microsoft Management Summit.
In the short term it clearly the strategy of the Microsoft Hyper-V team to position VMware as an â€śextra layerâ€ť (server virtualization is included in Windows Server for free â€“ why would you want an expensive and complicated extra layer). This might work for SMBâ€™s or SMEâ€™s who do not need all of the advanced VMware functionality, but it is not going to work for advanced business critical applications in large enterprises. In the long term, it is clearly the Microsoftâ€™s strategy to re-establish the central role of the Windows OS as a platform for server based applications by moving the virtualization layer up one level from below the OS to between the OS and the applications. The strategy here is to declare that VMware is irrelevant because you do not need to virtualize operating systems any more (since you can virtualize applications).
The clear winners here are enterprise customers who would really like to have some negotiating leverage with both Microsoft and VMware. Both of these vendors are determined to put the other out of business, yet both must work with each other to ensure interoperability for joint customers. This maximizes choice for customers and gives customers the badly needed leverage against both vendors in enterprise license agreement negotiations. As for who is going to win the war of survival â€“ sorry it is just too early to tell.