KVM in SUSE, what is going on at Novell?

As of Service Pack 1,  SUSE Linux Enterprise Server 11 (SLES) supports KVM. The bald facts are as follows.  SLES 11 SP1 is based on a 2.6.32 kernel and is now full supported on x86_64 processors which support hardware virtualization, for the following guest operating systems:

  • SUSE Linux Enterprise Server 11 SP1
  • SUSE Linux Enterprise Server 10 SP3
  • SUSE Linux Enterprise Server 9 SP4

We note there is no mention of other Linux guests or Windows guests. This post follows on from our previous post regarding the demise of Xen in Red Hat Enterprise Linux, and perhaps suggests the beginning of the end for Xen-based virtualization in Linux, but the story is far from clear.

It is impossible today to write an article about Novell without mentioning that the company is “up for sale” following an unsolicited approach from a private equity house.  It has revenues of around $840M and valuations are around $2Bn, however there is around $1Bn of cash, so the enterprise value is only 1.3x Revenues. Red Hat, by contrast is currently worth $5.4Bn on revenues of $740M, an enterprise value of 6.2x Revenues.  Some of this is difference is down to market sentiment around dominant players, but Novell seems a much more complex entity having a mixture of Linux and non-Linux technologies.  The private equity strategy seems to be to break Novell up so that the pieces are worth more than the whole.

SLES is appealing because it is actually growing quite strongly, possibly in underlying terms proportionally faster than RHEL.  But it is quite hard to understand the actual level of SLES subscription because of a complicated pre-payment deal whereby Microsoft licenses SLES for re-sale, and pays whether or not an end-customer subsequently licenses from Microsoft. This in turn maps through into a complex revenue recognition calculation for Novell, some of which is currently winding through into an artificially-reduced growth rate for SLES subscriptions (because historic SLES sales were bolstered by “shelf-ware” purchases made by Microsoft on behalf of non-existent customers).

The focus of this Novell/Microsoft deal, struck as long ago as 2006  and still to run until 2012, is predominantly virtualization. It is an initiative to ensure SLES and Windows play well together in virtualized environments in both directions, partly by providing optimized drivers.  Clearly in 2006 the deal was struck in the context of Xen, and given the similarities between Xen and Hyper-V, collaboration made perfect technical sense.  Now with the emergence of KVM, it is less clear where the synergies are.

However, although the exact terms of the deal are not public, it is likely that Novell cannot simply move from Xen to out-of-the-box KVM without breaching the terms of its “technical collaboration”. In which case it may forfeit some part of those revenues (and, given payment was made up front, this would likely trigger cash repayments back to Microsoft). In the meanwhile, although the term of the Novell/Microsoft technology development agreement was 5 years, the exclusivity of Novell as Microsoft’s Linux commercial partnership was only 3 years, so Microsoft is now free to develop its relationship with Red Hat and thereby get KVM guest support however it sees fit.

In this context Novell has a project called Alacrity VM which is described as a new “Hypervisor”, but is actually a set of enhancements to KVM to allow drivers to operate against devices present on a virtual bus, rather than against emulated device drivers in user space.  These devices are thus paravirtualized, but in a different way from the device level paravirtualized drivers in the standard Linux virtio libraries.  As well as showing performance enhancements for Linux on KVM, Novell has indicated it will develop Alacrity drivers for Windows, and this may provide a route to satisfy the letter of its agreement with Microsoft, whilst moving off Xen and onto KVM.

However, Novell will ultimately need to push the Alacrity KVM enhancements into KVM itself, and there is no sign yet that Linus Torvalds is onside. Novell has alienated the Open Source community through other elements to the Microsoft deal (in fact there were three separate deals) that introduced protection against infringing Microsoft patents for consumers of SLES and Mono (Novell’s re-implementation of .NET for non-Microsoft patforms) but only if you buy from Novell, not from other package maintainers.

There are doom-mongers who suggest that if Novell is acquired or split up, SLES would die.  The reality could be very different, just as EMC was able to liberate value when it separated VMware out, SLES may well be more valuable when packaged with a small set of compatible Linux technologies into a profitable growing company.

However, it probably wouldn’t stand alone for long.  One possible acquisition candidate is IBM, a long-term SLES partner.  IBM and Novell are tied together in opposing legal action from SCO regarding the alleged use of Unix IP in Linux, which would have the side-effect of allowing SCO to terminate IBM’s AIX licence. The acquisition of SuSE by Novell was facilitated by a $50M purchase of Novell shares by IBM, which we believe it still holds. Although the SCO case is definitely going towards Novell/IBM at the moment it is by no means settled, but IBM is unlikely to acquire SLES outright.  It doesn’t do GPL

Posted in SDDC & Hybrid CloudTagged , , , , , , , , ,