by Paul Rudo on 01/07/10 at 2:59 pm
The benefits of virtualization have already been stated here many times, so I won’t go into them again. It’s really no wonder that server consolidation has become one of the hottest business IT trends of all time.
Instead, I’d like to draw your attention to one of the challenges faced by virtualized environments.
IT professionals like to brag about what percentage of their systems are virtualized, and often envision themselves increasing this percentage in a linear manner… over time. But the reality is that there will be a plateau very early on in the migration process.
Usually, the first 20% of consolidation is comprised mainly of less critical systems such as DNS or test servers.
When it comes to the really demanding or critical systems such as databases and email, the challenge becomes much more difficult. In order to consolidate these systems, you need to think about future trends in resource usage.
- What would happen if 3 systems experienced a spike at the same time? This is a common occurrence during major events such as the Christmas Eve rush. Could you accurately predict the resources required to support such a spike? And what effect would this over-provisioning have on your ROI?
- What are the future trends for usage within your company? Has usage growth stagnated, or is it growing exponentially? How accurately can you predict your trends over the next week, month or year? Do you have visibility into these trends on a server-by-server basis?
- How far can you push your implementation before you risk violating your Service Level Agreement?
Part of the “magic” of virtualization is that it lets you exploit peak periods, and use them to your advantage. For example, email systems experience a major spike at 9:00 AM, then the load is much lower for the rest of the day.
In this case, a failure at 9:00 would be catastrophic. That’s why email servers are over-engineered to support a much higher load than they would ever be faced with.
There would be no point in consolidating your email server with another system that spikes at 9:00 am. You’ll just end up building a server that’s twice as over-engineered. It’s simply too much work for too little benefit. A better approach would be to virtualize your systems in order of ROI, starting with the ones that offered the greatest cost savings.
That’s why it pays to know what the loading trends are for your servers. Consolidation of your email server and backup systems would be a perfect synergy, since one spikes in the morning… and the other spikes in the evening.
But this example just covers a very simple example involving just 2 basic services. This problem becomes much more complicated when we’re dealing with 200 servers across 10 locations.
Before you can begin consolidating, you need to gain insight into resource usage. But how? What tools or techniques has your company been successfully using? I’d love to hear your comments.