Scott Lowe, while stating that he isn’t a Xen expert, has several issues with Xen, it seems. Not that he’s inherently anti-Xen, but honestly responding to the wealth of hyperbole that’s out there (from all sides). So his is not an attack, but a reasonable challenge and request for clarification. Since what I’m about to say is also relevant to some follow up to my previous postings about the enterprise readiness of Xen and XenServer, here is a nutshell version of why we are not simply trying to copy VMware, but deliver virtualization as a component of an application delivery stack. Before I respond, it’s important to note that though Scott phrases his comments in terms of Xen, they have nothing to do with Xen per se, but actually our product, XenServer. Indeed, some of the comments to my previous blog are about Xen, and not XenServer.
Also, before I address specifics, let’s just deal with the concept of “enterprise ready” once and for all. The largest deployment of virtualization on the planet uses Xen, and that is at Amazon. Is Amazon not an enterprise? Every enterprise has a unique set of requirements, so when someone says that XenServer is not enterprise ready, my response is to ask what specific features are missing, and to prioritize those for development. I specifically reject statements by VMware resellers that we are not enterprise ready just because we have a different form factor of the product. But I do accept that we don’t yet have all the bells and whistles that VMware has, and so that means that some people won’t buy our product. We certainly do have every feature that VMware had for its first $1BN of revenue. But here’s a good example of something we don’t yet have: certification against EMC Clariion storage arrays. Anyone care to guess why? Fortunately with the help of our ecosystem partners, and with the Citrix portfolio of add-on features, we have an incredibly compelling offering - one that is suitable for most enterprises and that addresses both native and virtualized workloads.
Memory: Scott points out that VMware supports up to 256 GB memory per host and says that we support less. Actually that’s not true. Our testing limit of 128GB/host ($40K of memory!) is just that. If someone wants to put more memory in the box, will the system work? Yes. The architecture scales directly to 4TB. Does anyone care? That would be a very expensive box. I’m not sure yet, but likely XenDesktop will exercise a lot of host memory in some implementations.
Paravirtualization: Scott states that VMware also supports paravirtualization, and also points out that paravirtualization is useless for legacy guests. Yes, VMware with the vmxnet driver supports what we’d call a PV driver. So, rather than decry Xen for equivalency, Scott should acknowledge that Xen led the industry in development of paravirtualization, and helping to optimize OSes for virtualization (including, through XenSource and now Citrix) all future Windows kernels, which are “enlightened”. We’ve always argued that the basic engine over time should be (a) equivalent (arguably identical and compatible) and (b) commoditized, in all offerings. So Scott, we take your arguments as a powerful argument for Xen’s innovation and creativity, and acknowledge your thanks to the Xen community for helping VMware to find a better model for the future development of its hypervisor. I think I was the first to welcome VMware to the era of Paravirtualization.
Where we (XenServer) do have a very significant edge, and Scott doesn’t have all the facts is that .. with some careful work one can really optimize the performance of legacy Windows on Xen’s PV hypervisor – which we’ve done. So we do a very good job for Windows virtualization. Not all Xen implementations do, by the way – we regard that as proprietary value add. I also repeatedly hear about lack of HA etc in XenServer. Well, I found a customer using NetScaler as an application layer HA tool for XenServer recently, and this is a perfect example of why XenServer, with the Citrix portfolio, offers a richer environment for customers than simply jamming every known IT infrastructure function into the virtual infrastructure layer. It is also true that XenServer is still a follower to VMware VI3. Hey, they’ve been at it 10 years and we’ve been going for 3. But at a core functionality level, we are absolutely there, and the relative differences are rapidly diminishing. Working with a rich ecosystem of partners is the right way to finish off the remaining gaps.
At a fundamental level there is a key difference between the way we build our product and the way it is factored for market, versus VMware’s: We have an open, pluggable storage architecture with no clusterfs, but support a range of storage back-ends including tight coupling to virtualization-aware storage infrastructure, such as NetApp arrays and the ONTAP API. We have a powerful partnership with Symantec, and there are more storage vendors to come. We like it, storage vendors like it, customers like it because it doesn’t break existing storage management procedures, values the contribution of the storage infrastructure vendors, and fits in with what customers have today. We don’t have a centralized management console that purports to be the god of all things virtual, and therefore represents a single point of failure. We do have a redundant management architecture that explicitly endorses 3rd party plugins, and we go to market with a strong ecosystem of ISV add-ons, rather than excluding them from the opportunity. And we’re not a systems management vendor, so we’re delighted to plug into the management console of choice of our customer – System Center, Director/Tivoli, HP VMM/OV, BMC, CA – you name it. We haven’t ever shipped a hotfix, and VMware ships more than one per week. In my view, the patch rate is a very significant statement about the failure to meet enterprise quality levels on the part of VMware. I’m not saying we have no issues, just that we haven’t needed to ship any patches yet – and we’ve had the ability based on our relatively smaller footprint, to deliver 2-3 releases per year.
We offer much better price performance, and a portfolio of products that extends the value prop of virtualization up the app delivery stack, and through the network itself. From an end to end perspective, we have more ‘stuff’ to make the user’s use case work than VMware does, and therefore have a very powerful set of offerings.
Finally, we offer solutions that address both native workloads and virtualized workloads – through XenServer Platinum which includes dynamic provisioning for both native and virtualized workloads, “any-ness”, and comprehensive workload management while dramatically reducing VM / image sprawl. VMware does nothing for the 90% of servers that today are not virtualized, other than promise to virtualize them for an exhorbitant fee.
That’s about it.