Posted by Fraser Campbell
Thu, 26 Jan 2006 12:25:00 GMT
Not entirely a Linux story but we love Xen so why not …
It has long been known that Sun planned to virtualize their Opteron boxes using Xen, see this Nov 2005 release for example.
Recently they are making noises about officially supporting Xen on Sparc as well. Xen on a T2000 (64 thread CPU) would be interesting but the question remaining is whether a few threads of a T2000 would give a virtual machine sufficient performance to run at more than a snails pace.
Read more...
Posted in Xen | no comments | no trackbacks
Posted by Fraser Campbell
Sat, 14 Jan 2006 22:13:00 GMT
The Xensource wiki now contains a very short Xen Hardware Compatibility List.
Hopefully this develops into a useful resource, currently only a single HP server type is listed. Please contribute your hardware configurations to build up a useful list of known good and bad Xen hardware options.
Posted in Xen | no comments | no trackbacks
Posted by Fraser Campbell
Wed, 04 Jan 2006 00:27:00 GMT
Redhat employee Daniel Veillard has announced “Libvir the virtualization API”.
The goal seems to be creating a stable C API that might eventually be hypervisor agnostic (currently only the Xen hypervisor is supported).
Python bindings are being worked on, let’s hope for Ruby bindings soon.
See Libvir homepage for details.
Posted in Xen | no comments | no trackbacks
Posted by Fraser Campbell
Fri, 23 Dec 2005 14:05:00 GMT
Kris Buytaert has an excellent article on automating Xen infrastructure using a number of open source tools such as SystemImager. See the article at
http://howto.x-tend.be/AutomatingVirtualMachineDeployment/.
This article should give you many things to think about.
Posted in Xen | no comments | no trackbacks
Posted by Fraser Campbell
Wed, 21 Dec 2005 23:34:00 GMT
Xen 3.0 is available for general use but there are a few major shortcomings for a typical Linux administrator:
- patches that are necessary for Xen functionality are not available in the standard kernel
- there are no good high level tools available (not that we have any proof of at least)
I cannot address the first issue, I can only hope that Xensource and Redhat do a good job of reworking their patches and getting them into the mainline kernel.
The second issue, I have decided, I can do something about. After poking around in the Xen libraries and perhaps inhaling too much polluted Toronto air, I have decided I will write the high level tool that people need.
The project will be released under an open source license (undecided on which license so far) and it will be called GOXen. The GO portion of the name was chosen because the software will be mostly written while travelling on a GO Train between Toronto and Georgetown (Ontario, Canada).
The heavy lifting for interfacing with Xen is done already in the open source product. LGPL licensed libraries in C and python are available. The key component missing for a usable Xen environment is the high level interface suitable for multi-server use.
I will document my trials and tribulations on the GOXen website, as soon as useable code is available it will be released for public consumption.
Posted in Xen | no comments | no trackbacks
Posted by Fraser Campbell
Wed, 21 Dec 2005 04:24:00 GMT
Ziff Davis author Steven J. Vaughan-Nichols wrote an opinion piece recently stating that Xen is not the next killer app. It is refreshing to see something other than an article which resembles a Xensource corporate press release.
Xen is a very powerful piece of software, it is a very useful and important piece of software but it is evolutionary, not revolutionary. Xen appears to bring large performance improvements to the x86 virtualization game but the end-user experience isn’t significantly different with Xen than with other virtualization solutions (VMWare’s ESX for example).
Xen will gain steam through 2006 but until it is integrated with the mainline kernel (the kernel from kernel.org) it is likely to remain a small player.
If we see the Xen architecture integrated with mainline soon (we can always hope), then Xen deployments will grow like wildfire.
I believe Xen will remain a Linux/BSD/Solaris only solution for at least 2-3 years, if Windows proves to work well on VT-x capable hardware then Xen could eventually become a major player.
Yet another possibility is that Microsoft releases a Windows port to Xen and users can take advantage of paravirtualization even on Windows. A Windows port to earlier versions of Xen was mostly complete so I expect the job would be an easy enough. However, chances are good that Microsoft won’t want to touch an open source hypervisor.
If you really want to find the next killer app (open source or not) I think we have it, it’s Ruby on Rails.
Posted in Xen | no comments | no trackbacks
Posted by Fraser Campbell
Mon, 05 Dec 2005 15:30:00 GMT
Xensource today announced the release of Xen 3.0.0. This is the first major release of Xen in over a year, the release focuses on features required for enterprise readiness:
- support for SMP guests (up to 32way)
- PAE support extending 32bit address space beyond 4GB
- up to 1TB of memory on 64bit systems (Xeon or Opteron)
- itanium support
This release of Xen saw contributions from all corners of the IT industry—AMD, HP, IBM, Intel, Netapp, etc. It seems that an open-source hypervisor standard is highly desired by most hardware vendors.
Xensource claims near-native performance for virtualized operating systems
See Press Release for full information.
Posted in Xen | no comments | no trackbacks
