I read recently a very well written article about Data Center Utilization and what is actually hidden behind this information.
Before we actually start bashing the different Managers about the CPU Cycles – there is a point, it’s more about how efficient we utilize the services we have in place. CPU Cycles is only a part of this, we also need to look at numbers of Clients/Connections/Sessions that we are utilizing – and at the same time how we we serve our clients.
As taken from the article: CPU isn’t the only resource when it comes to determining the effectiveness of a server.” ………. took a look at the utilization rates of about 1,000 Mozilla servers. Here’s what he found: the average CPU utilization rate was 6 percent; memory utilization was 80 percent; network I/O utilization was 42 percent.
When measuring performance, it is clearly “In with a lot of iron and we can serve an amazing number of clients, in microseconds this and that. … But my questions is also – is this really relevant.
Are we sometime forgetting about reality – we do not serve a gazillion sessions, a few hounded (thousands) is sufficient and we are home safe!!!!.
Here is where I suppose that people and companies have to stop op and think: Are we now facilitating an unrealistic number (of clients) just to keep our Job/Importance or something like this; should we rather think about how we can serve the business and scale up (when needed).
I have seen many clients (and colleges) working on Internal (Private Cloud) Offerings, where we really should have utilized the PAAS/SAAS Offerings, there is an abundance of these out there.
When I look around to find Amazon EC2 (http://aws.amazon.com/ec2), Microsoft Azure (http://www.windowsazure.com) and Google Services compete for the offerings, I think many customers should investigate if this could be a valid contribution to their internal (existing) infrastructure.
Who would dare to break the trend – take a “mixed approach” and truly investigate if the need is for a Private Cloud – or a Public (Access Controlled Cloud) and hopefully secure better utilization and save on the CAPEX in the end.