Spouting Shite

Upscaling servers makes sense, but upscaling a system is better.

One of the great problems that any system manager has to deal with is capacity. In a strange way we all aspire to have capacity problems…in my opinion capacity problems come hand in hand with success and it is very rare that you have one without the other.

If you work for an enterprise then the problem is rather academic…you simply throw money at the problem. Well, we don’t live in the perfect world. I run a small business and our system has evolved over a period of two years under my ‘bootstrap and transition’ business plan so we have a very diverse network full of equipment from various brands and technologies….when we think of the emailcloud network we don’t think of homogeny

So, how does a system designer / manager deal with capacity issues? Well…there are a few rules:

1. Downtime is unacceptable
2. 7 P’s (Prior Planning and Preparation Prevents Piss Poor Performance)
3. Listen to your friends but emulate your peers
   All too often we make strategic infrastructure or technology decisions on a whim or on the advice of well intentioned friends or colleagues….often these are the first steps down a very long dead end.

In our case we deliberated for months about how to add capacity to our infrastructure. After years of buying bulk equipment on ebay and cabling it up by the cabinet load we came to a number of pinch points:

1. Power. The average age of our infrastructure components were 3.4 years old. With the advances of new power units we felt that we could save on our power bills significantly by replacing some of our equipment. But, buying new servers is expensive, event if you use the new Carbon Trust 0% loans
2. Upscaling servers makes sense, initially. Well, buy being more specific about what servers you buy you could have an investment strategy to upscale them over a period by adding components…for example adding more RAM or upgrading the disk drives to faster models etc. These changes can be made very easily, but it is expensive and very soon you reach the limit of the hardware.
3. Re-design the infrastructure / system….this is a never ending process.

I believe the answer to adding more capacity (on a budget) is to use all of the above, the mantra should be ‘upscaling the system’ rather than ‘upscaling servers’ makes sense.

Ps. I hope that this article kicks the ass of the other GoogleGroope hopefuls