Whether  you are looking at setting up a new network or adding a new server to an existing infrastructure, the process of sizing, scaling, costing and finally purchasing a new server can be pretty daunting.

The role a server will play in your networking environment greatly affects the specifications of the hardware involved. Of course, the hardware configuration of the server will have a direct impact on the cost too. It is not uncommon to make a compromise between cost and functionality in this day and age, but what you need to ask yourself is whether you are making an informed compromise or just simple cost cutting? The one can provide you with an effective solution where the other can land you in serious trouble.

I cannot count the number of times I have walked into a client’s site to solve some problem only to arrive there and not be able to help since their hardware is under specked and over loaded. In this scenario, my only recommendation can be to purchase additional hardware, which in most cases is just not on the cards so they have to live with what they got and work the cost into their next budget, and in the mean time put up with unhappy end users.

Now some good planning on your part, as the IT Manager, Administrator or Service Provider can avoid these kinds of complications, increase the value to the client and ultimately lower the total cost of ownership.

So what are should we consider when purchasing a new server?

1.       The role of the server: Different server roles have different hardware requirements. Database servers for example place an emphasis on Disk IOPS (input/output per second) where Web Servers prefer more Processing power. Multiple roles may require a combination of multiple requirements, its up to you to decide the roles and stick to your decision once there.

2.       Availability: Some servers are considered more critical in terms of workflow operations on a daily basis where other server roles tend to be redundant in one way or another, Email servers, for example are usually considered mission critical where Domain Controllers are normally installed in pairs or more, making them redundant. The company would continue to function should one of your domain controllers crash, but should one of your exchange servers crash, all the user’s with mailboxes on that server would be unable to send / receive mail until you have recovered from the disaster.

3.       Redundancy: Redundant power supplies are a must for every server today, redundant hard disks a definite consideration, the level depends on the role of the server and data stored on the array, for DC’s the norm is 2 disks in Raid 1 or Raid 1+0 (Raid10), for exchange servers the recommendation is system files & Logs on Raid 1 or 1+0 and information stores or databases on Raid 5 for higher reliability and performance (Disk IOPS)

4.       Scalability / Expansion: It is the nature of networks to grow, software to be upgraded, etc. The average lifecycle of a server is 5-10 years, you need to take into consideration paths and methods to deal with the projected changes over through out the lifetime of the server, this should include upgrades & purchasing of additional servers & licenses.

5.       Backup: Backup and Restore plays is a major part of any network and you should plan accordingly from the start. Ask yourself how much information from each server needs to be backed up, what is the acceptable time frame for the backup to run, what is the acceptable time frame for a recovery / restore process, what are your licensing requirements and finally whether a network backup or local backup should be performed to meet these objectives.

6.       After Sales Service and Support: The after sales service and support is a huge factor in the availability and functionality of your server, so always consider such things as “is the hardware locally supported”, “what is included in the warranty”, “is there a Service Level Agreement available and what benefits does it have”, “in the event of hardware failure, how quickly can I obtain a replacement part, either from the supplier or another independent vendor”. These are important questions since some suppliers will sell you hardware and claim support for it then in 6 months time when the system board fails, you get told they need to order one and it can take up to 6 weeks to deliver.

7.       Cost:  Finally you should consider cost, I have left this till last because in my opinion when it comes to deciding on your infrastructure this should be the last thing on your mind. Yes, it is a deciding factor, but just as important is the question of how cutting cost will effect the availability and performance of your systems overall. A compromise between functionality and cost is always required but rather than placing emphasis on cost, one should place emphasis on availability and functionality, at the end of the day, the server hosts the business applications, without its services (or if it underperforms) most companies become severely crippled or in some cases unable to function at all and someone’s going to pay, are you going to let that someone be you?

Sales people will always overwhelm you with quotations, proposals and recommendations, and persistently harass you for that order. The main thing to always bear in mind is that most sales people are just that, sales people, and are not trained or geared up to correctly spec the server to suit your specific needs. As such, the onus falls on you to research your requirements and then veto the quotation.

There are various tools out there on the web which will help you scale your servers for particular applications and roles, my advice here, find them & use them, it WILL save your bacon in the near future.