188
Networking: A Beginner's Guide
you needed to spend. Under-purchasing leads to additional, unplanned purchases,
which might include adding more disks or more memory, or even needing to replace
the server much too soon. Over-purchasing means you spent more for a server than
necessary, which might lead your company to deny your request for a particular server.
Instead, you need to find the "sweet spot" for specifying just the right server for your
needs; then you can defend your required configuration and its cost. You can't do any
of this unless you have clearly defined your needs.
To specify the needs for a server clearly, you must be able to answer all the following
questions:
What is the useful life of the server? How long do you expect to use the server?
Will you replace it in two, three, or four years? (Most servers are used for
around three years before being replaced.) Everyone should agree on this time
frame, because if you plan to replace the server in two years, you can get by
with a smaller server than if you need one to last three or four years. If you
specified a server capable of meeting two years' needs, however, you don't
want to get to the end of two years and then find out that your company won't
approve a replacement.
What job will the server perform? Will it be a file and print server, a web server,
a database server, or some other kind of server?
How many users will the server support and what are the needs of those users?
For example, with a file and print server, you must estimate the storage and
bandwidth requirements needed to satisfy all the planned users' requests. For
a database server, you must know how quickly the server needs to respond to
various database operations.
How reliable must the server be? What are the consequences (costs and impacts)
if the server crashes for one or more hours, or for a day or two?
Will you use clustering for the server? Clustering is a technique whereby
multiple servers share the same essential job. If one fails, everything keeps
working, albeit at a slower rate. Once the failed server is repaired, it can then
be added back to the cluster.
How safe must the data on the server be from loss? This is different from the
preceding question because you might have cases in which a server must never
lose data, even if it isn't a big deal if the server goes down for a few hours. In
such a situation, you would use a RAID 1 or RAID 10 configuration, but you
might not care too much about, say, redundant power supplies. You might also
explore some kind of hierarchical storage scheme, where data is automatically
copied to tape or optical disk in real time, or where you make several live
incremental backups of files each day.
If the server fails, what are your backup plans? Do you plan to keep a hot-spare
server (one that's ready to be swapped in at a moment's notice for a failed
server) available or do you plan simply to rely on the server manufacturer's