This is the second part of my look in to virtualisation, its taken a while to get here as life got in the way.
Looking through System Center Virtual Machine Manager right now I can see we currently have 16 Hyper-V powered virtual servers in production, with a hand ful of other VM's going through testing to reach production. Looking back it has taken us some hard work to get where we are currently and in getting here we've learned serveral important lessons regarding Virtualisation and how you handle it.
Our Virtualisation Platform of choice OS Microsoft Hyper-V we currently run 3 production Hyper-V hosts running Windows Server 2008.
In the past we have investigated other platforms, starting with our first foray in to the world of Virtualisation with Microsoft Virtual Server 2005. We have also used the Parallels Virtuozzo product unfortunately several failures with the system left a sour taste in our mouth, we retain one unsupported Virtuozzo server for legacy VM's which can't be migrated due to time and money issues (IE: There is neither available right now!).
One question which pops up repeatedly is why we don't run VMWare, the simple answer to that question is the price of the software. At the time we were investigating possible Virtualisation software VMWares cost was just too high for what we had hopped to achieve.
The choice to go with Hyper-V came down to two things, firstly the price and secondly as my boss put it the 80/20 statement. Hyper-V would do 80% of what we needed 100% of the time, and as we can afford to live without the other 20% its perfect for our needs.
With the road map for Hyper-V I cannot see us moving away from it as a Virtualisation platform, Live migration in Hyper-V R2 and the forthcoming ability to over commit memory will go along way towards providing that other 20% of the functionality.
The shared storage behind our Virtualisation platform is a Dell / EMC AX150i SAN, as an entry level device running smaller workloads the AX150i works okay, the administrative interface is dog slow and the chassis supports 7200K RPM SATA drives. In its current configuration our AX150i is capable of approximately 500 IOPS, this however falls short of the kind of performance we find ourselves needing. During an average day the SAN now needs to maintain at least 1000 IOPS with a peak capacity of double that for when some of our more intensive tasks kick in. After discussions with Dell its become apparent that we can increase the performace of our AX150 by replacing all of the disks and buying a second disk chassis, however the cost of doing this is rapidly approaching the cost of the SAN in the first place.
Knowing where we are has lead us back to Dell who are preparing to undertake a Virtualisation Readiness Assessment on our servers, they will be monitoring all our systems over a 14 day period to gauge exactly what our workload and provide us with some recommendations based on those numbers. Whether those recommendations are to virtualise more systems or to stick with physical servers remains to be seen, and verified by a third party.
You need to be a member of The SysAdmin Network to add comments!
Join The SysAdmin Network