Hypervisor Visionary

Qumranet's CEO Benny Schnaider discusses his company's open source technology and the future of the virtualization market.

Benny Schnaider is a man who figures to have some influence in how competition in the virtualization market plays out. Schnaider is the CEO and co-founder of Sunnyvale, Ca.-based Qumranet Inc., which is the sponsor of the Kernel-based Virtual Machine (KVM) technology. KVM is an open source alternative to the Xen hypervisor, which recently became an approved component of the Linux kernel.

Endorsed by Red Hat Inc. and Canonical Ltd.'s Ubuntu -- as well as supported in hardware by Intel Corp., Advanced Micro Devices Inc. and IBM Corp. -- KVM handles both Linux and Windows guest operating systems without preference.

Schnaider sat down with Redmond Editor Ed Scannell to discuss the strategic role KVM might play in the virtualization market, as well as his views on how he sees the battle between Microsoft and VMware Inc. shaping up.

Redmond: What strategic role does KVM play in virtualization?
Schnaider: You have to look at it from two perspectives: closed source and open source. There will be a big battle between VMware and Microsoft on the closed source front, and there will be a battle between KVM and Xen. I think VMware will gravitate toward open source, but I don't think Microsoft will go into that area at all. Our product is based on the Linux kernel but it can run Windows [operating systems] better than Linux. We'll also have an offering for hybrid environments with both Windows and Linux. I think our advantage here is the simplicity of the solution and the alignment of KVM's architecture with the open source mentality.

What's technically different in KVM?
Understand that a hypervisor is an operating system of operating systems. So from the start we were not going to invent a new operating system complete with things like memory managers, I/O subsystems, schedulers and drivers. So, what this allows us to do is, if Intel comes out with Symmetrical Multi-Processing we can get things from the community and make adjustments for that in our product. We don't have to write it ourselves. This benefits the [development] community and end users because they don't have to learn new commands.

We took an existing OS and added a small amount of code that makes the OS a hypervisor. Our line count -- with KVM -- is around 20,000, compared to Xen, which is about 300,000. Small in this world is always better.

Are you worried about getting a late start in this market?
Yes, but there are advantages to starting late. When we started KVM, virtualization support was already baked into the microprocessors. This was not the case for VMware ESX or for Xen. We were able to build an architecture that's more modern and simpler because it's based on reusing the existing hardware, making it a lighter-weight hypervisor.

"KVM can run on a rich set of architectures including the x86, IBM's 390 mainframes, the Power PC, Itanium and eventually on embedded processors. If you seriously believe in virtualization being everywhere, then it has to run on every architecture."
Benny Schnaider, CEO and Co-Founder, Qumranet Inc.

So you feel KVM is gaining some real traction against Xen?
Yes, and I think it has to do with ease of development and maintenance. The number of people you need to maintain KVM is probably one order of magnitude less than the number needed to maintain Xen. Another reason is KVM can run on a rich set of architectures including the x86, IBM's 390 mainframes, the Power PC, Itanium and eventually on embedded processors. If you seriously believe in virtualization being everywhere, then it has to run on every architecture.

Will hypervisors be a standard component of both closed and open source operating systems?
VMware and Xen tell you the hypervisor should be a separate layer from the OS, and then companies like Microsoft and Novell tell you the hypervisor should be part of the OS. If I have to predict over the next few years, [I'd say] it will be part of the OS in both the closed and open source worlds.

But every time Microsoft bundles something into the OS, entire sub-industries disappear.

Will this happen to the virtualization market?
Yes, I do think so, and I think one competitor on the line is VMware. Microsoft will continue to add features to its servers while keeping the price identical with previous versions, and virtualization will be one of those features. Microsoft will eventually improve it to where VMware will have a very hard time competing. [VMware] won't be able to charge $5,000 for their hypervisor anymore. It'll be a significantly lower number.

Also making it difficult for companies like VMware will be the commoditization of virtualization through things like more features built into silicon, and open source products like Xen and KVM.

So what does VMware do?
If you follow their strategy and recent announcements, it's not all about virtualization. It's about app management and virtualization management. [VMware] acquired B-hive [Networks Inc.] recently, which does performance management. They're trying to get out of the pure virtualization play and into something that creates value on top of virtualization.

Two years ago we said creating another hypervisor in a closed system environment wouldn't be a smart thing to do. It's one of the reasons why we open sourced it and are focusing on other things.

Can VMware protect itself by developing a robust third-party market for its infrastructure products?
I think you're right. VMware is a great company with great products, and Microsoft is known for taking more than one iteration to get things right. So the runway they have is longer than people think. But eventually Microsoft will get it. But one thing I'm sure of: VMware can't stay at the hypervisor level. That's a suicidal move for them.

About the Author

Ed Scannell is the editor of Redmond magazine.