Is server virtualization delivering for you yet?
Real world, real benefits?
Reader Workshop In one of our recent Tech Panel surveys, conducted towards the end of 2008, we asked respondents to call out what they felt were the most positive trends in the IT industry during the previous 12 months.
In answer to this totally unprompted question, the word ‘Virtualization’ came up more than four times as often as any other significant word or phrase, including terms such as ‘Cloud Computing’, ‘Open Source’, ‘Mobility’, and even ‘Price Cuts’.
This positive view of virtualization was confirmed in a subsequent survey conducted just a couple of months ago in which 53 per cent of the Reg readers who responded said their organisation had embraced the approach, and a further 19 per cent said they were considering it.
While such sentiments are very encouraging, to what degree is virtualization — particularly based on popular technologies aimed at x86 server platforms — delivering on the promise?
Indications are that most large organisations and many smaller ones have now gathered experience in this area as part of their drive for operational efficiency. Server consolidation has been the big win here, with claims of an order of magnitude reduction in the number of servers being frequently heard. Such reductions in server hardware requirements, if they are true, can not only save money on kit, but in theory lead to space savings, reduced power consumption and lower operational overheads from a maintenance and management perspective.
But how real are these claims and benefits?
The chances are that if you are reading this, you will have some experience yourself in deploying one or more virtualization solutions, and to kick off this latest Reader workshop, we’d be interested to hear how it’s been for you.
In general, experiences thus far seem to have been positive, with virtualization delivering on the key benefit being advertised — namely, reducing hardware costs. There are a number of quite straightforward scenarios in which virtualization can be adopted, such as enabling the provision of development, integration and test environments, supporting the consolidation of less onerous workloads onto a reduced number of servers, giving organisations the ability to make use of unoccupied server resource such as warm standby machines and disaster recovery sites, or indeed a combination of all three.
But where are such benefits being realised specifically? With regard to consolidation, for example, what kind of ratios have you achieved — 3:1, 5:1, 10:1, higher? And how does this vary by workload? Have some jobs or applications given rise to better results than others? If so, why, and what advice would you give to someone figuring out where to start or focus their efforts to get the best returns?
It would also be useful to hear some of your views on the enabling technology itself. Some have expressed concerns, for example, about the expense and maturity of solutions, particularly as you scale up your deployments or start mixing and matching different operating systems. Have you encountered issues in these areas, either generically or with specific vendors or products? Do you have any tips or workarounds to avoid cost escalation or to overcome technology limitations? Are skills and experience the answer, and if so, how easy are they to come by?
We’ll be drilling into a lot of these areas over the coming few weeks as our workshop unfolds, but if you’d like to kick us off with some high level views of the practicality rather than the promise of server virtualization, join the debate by adding your comment below. ®
Intel IT's experience w/Virtualization
As this is my first blog on this forum, I'd like to introduce myself. My name is Bill Sunderland and I have been working at Intel for 13 years primarily working on Server Hardware Engineering and the last four years of which I have focused my efforts on Program Managing the Virtualization Engineering releases for Intel IT. Last year, I published a WP demonstrating the methodology we used:
While this WP is a year old now, it s still very useful to learn how a large corporation implemented virtualization. The challenges remain 'virtually' the same. As of 2H '09, we have over 1400 VM's deployed across eight major sites worldwide. We are currently on average at 9:1 consolidation ratios with calculated available capacity to achieve 15:1 ratios.
A superb new bit of information is our WP on Nehalem's performance. Here's the WP introduction and link: "Learn about Intel IT’s proof-of-concept testing and total cost of ownership (TCO) analysis to assess the virtualization capabilities of Intel® Xeon® processor 5500 series. Our results show that, compared with the previous server generation, two-socket servers based on Intel Xeon processor 5500 series can support approximately 2x as many VMs for the same TCO."
As you will see from the WP, we are anxious to begin our deployments on Nehalem.
There is one more good source of information we have on this topic in presentation format that is currently undergoing legal/proofing before being shared externally. This is due for release in 2weeks. However, i did want to point out a great key learning from the presentation with regards to our comprehensive ROI study we conducted. In short, we analyzed the sources of positive/negative ROI as percentages. 'Our study' showed the following results for deploying via virtualization as compared to normal physical servers:
40% Server Capital Reductions
11% Ethernet Switch Port Reductions
10% Power Reductions
7% Rack Reductions
3% Ethernet Cable Reductions
3% Ethernet Cable Run Reductions
3% Deployment Labor Reductions
10% Virtualization License Cost
9% SAN Cost
2% FC Switch Port Cost
This analysis really helps to understand where the savings are costs are coming from. As a result we looked at the negative ROI factors and tried to determine if we could reduce these costs even further. We found one opportunity in the SAN cost area and already implemented the change. We saw that we were using expensive FC HDD's for our entire virtualized environment. We noted that 70% of our virtualized environment was pre-production while the other 30% was production. As a result we determined that we could use the less expensive SATA HDD's for our pre-production environment.
I hope this information is helpful!
Intel IT Server Virtualization Program Mgr.
For them but not me
I used to work at a huge online betting company based in England, and we didn't use VMWare that much, a few things here and there but overall, virtulisation wouldn't have worked. The web server farms (3 sets of 25 blade web servers in each set) where really going some on a weekend, and these where beasts, database servers top of the range 64 CPU HP half a rack high bad boys really fucking going some.
VM wouldn't have worked, by the time we would have specced out a server to host VM's including future predictions, it was out of date. We where upgrading, installing upscalling servers all the time.
This being said however, I now work at a much smaller company and I am looking to install some virtulisation there as it would massively improve their services, HA, DR and my stress levels. We have a lot of servers at low power, but need seperate servers. When this has all be rolled out, high initial cost, but savings in the long run and a massively better infrastructure.
Different horses for Different courses I think the saying goes
Our experience in virtualization
I work in Intel IT. We have been studying the benefits of virtualizatoin and have start deploying it at a larger scale since last year. We were planning for a 10:1 to 16:1 consolidation ratio. I believe we are targeting 14:1 to 20:1. My colleague has documented our experience and lesson learnt in a white paper. It is an interesting read. You can find it here: http://communities.intel.com/docs/DOC-1513