Friday, 26 November 2010

V for Virtual

For some time now, we have been looking at a project to implement virtualisation. I decided that this would make for some interesting blog entries, and I thought that I might focus on this for a while.

First of all, I suppose that I should go right back to the beginning to explain some of the reasons behind the decision. When I first joined the company, the servers were mostly tower models, that were stood on a table in a small room. These devices had limited processing power, low memory and disk storage even by the standards of the day, and were not really up to the task required of them. It should be said that they were most definitely not cheap, but certainly could not be described as being good value for money.

It was identified that we needed to buy some newer machines to replace this old equipment and as a matter of some priority in order to provide urgently needed resources. As part of the project, it was agreed that we would move to rack mounted equipment; this made far better use of the available floor space, we could get a lot more in the same area. The equipment was not totally top of the range, but was very good quality, a good specification and thanks to some quite keen negotiation (though I say it myself) was pretty good value for the money.

This made a huge difference to operations. Within a short time, staff could see significant improvements in speed of operation, we had much better storage facilities, and it was all much more flexible. This all helped demonstrate that the investment was appropriate; and I was also able to confirm some of the benefits using some standard metrics.

But that was some 5 years ago. That same equipment is still functioning, and thanks to some upgrades is still providing a good level of service. However, it has been identified that across the estate, much of the processing power is underutilised. Although some machines made full use of their memory, more than half do not. We have a couple of servers with disks getting quite full, but the rest are using less than a quarter of the available space. The most obvious excess is in the network cards; generally, they are using less than 5% of the available capacity.

It was also identified that the specific servers were manufactured before the newer energy saving devices now available; they use quite a lot of electric power, both to operate and to cool. We ran some tests and found that they would operate just as well at a warmer temperature than had previously been used, and this helped to reduce the need for cooling, so it did save some electricity, but we felt that it should be possible to do better.

Of course, it was also identified that with equipment getting 5 years old, there was an increasing chance that we would see some hardware failure. This was the main concern for me; it seems foolish to be miserly with spending on hardware, when a failure could cause huge losses to the business due to loss of data or operational capacity.

After identifying the need for replacement equipment, we started to look at newer versions of the same hardware; this had a number of green options for power saving, but I was still concerned that we would be paying for extra capacity that never got used, even allowing for growth within the business.

Like a lot of people, I'd heard about virtualisation, but wasn't sure if it would really work for us. I was offered the chance to see some Dell kit in action, along with the Equalogic SAN units. These were really impressive, and gave a lot of options. I also compared these to some HP hardware with StorageWorks; these looked a little better if also a bit more expensive.

The next step was to consider what virtualisation software to use. I had some spare hardware and installed evaluation copies of both Hyper-V and VMWare. I also took the opportunity to see some Citrix systems in action. It wasn't really possible to do as a full a test as I would have liked due to pressure of work, but it soon became clear that the decision would come down between Hyper-V and VMware. I liked both and felt that either could do a really good job; it was just a case of which we felt we would be happier with in the long run.

At this stage, I managed to get some basic technical books for the two software products; I had hoped that this would help to make the decision a bit easier, but unfortunately, it didn't really help at all. In the end, I decided that we would go with VMWare; the product looked a bit more polished, it's been around longer and is more mature.

So at that point, I started to do some negotiation with the suppliers. This went on for a while, and yes, I played them off against each other. But ultimately, I managed to get a deal that I thought was worthwhile, that the supplier was happy with and that I could sell to the senior managers. There was a slight delay getting the stuff onsite, but it's all here now, and we are starting to install it; but that's going to be the topic for another occasion.

No comments:

Post a Comment