Following my last post I thought to clarify our position regarding Dell. I think they make great products which are cheap particularly in the desktop market. I think their servers are fine too. We did deploy a large Dell solution a couple of years ago and Dell contacted us from the US and consequently did an international case study: http://www.dell.com/downloads/global/casestudies/2006_smartyhost.pdf

The total solution cost us a few hundred thousand dollars. Importantly, not only did we deploy large servers but also an EMC SAN. A SAN is a Storage Area Network with a high-speed network of storage devices. The SAN contains only Fibre storage (not SATA). The benefit of the SAN is our customers’ information is stored and replicated across hundreds of redundant fibre disks. We don’t store any customer information on servers! You will find most of our competitors keep customer information on the same server they host on. Yes, they all use RAID but there is a single point of failure if the server fails or if there is corruption in the disks. The SAN can also perform round-the-clock data backup and restore tasks at high speeds without involving servers which is another real benefit. It should be noted that what we have done is nothing new. If you ask any Infrastructure Manager or CTO of large companies they will all be using SAN to manage and protect their data.

So why leave Dell? It is not that Dell is neither bad nor good. It is about matching the application to the right vendor. We wanted to keep our model of clustered servers connected to SAN but needed a more enterprise solution. We wanted more redundancy, more fibre, more clustering and we found SUN Microsystems’ (www.sun.com) solution was a better fit for us at this moment in time. One area where it has proved to be very beneficial is our new Vigabyte solution. I am still amazed by our competitors who offer virtual servers crammed onto a single server with no SAN storage. Some of the limitations of not having a SAN are:
1. Single point of failure - with no clustering of physical servers means if the server dies then so does your data.
2. With no SAN there will be disk I/O issues which will affect performance.
3. You cannot scale the storage as you are limited by the amount of storage crammed into the single physical server. SmartyHost has two SAN’s with capacity of one quarter of a petabyte! (1 petabyte = 1,000,000,000,000,000 bytes)

Oh, and if you are wondering what F5 BIGIP’s are in the DELL case study these are load balancers which we still use. We purchased two of these at $45,000 each. They also have a compression module which means we are able to load our customer’s website faster (up to 8 times) and save you on bandwidth at the same time. Qantas has the same model BIGIP as SmartyHost and has an interesting use for it other than clustering. Depending on your Qantas membership level BIGIP will direct you to either the slow or fast network to access information on their website.