Different ways exist to improve the availability of IT systems in the enterprise. But many times the best moves seem to be a little far from our reach because of the legacy infrastructure we have. The first point in planning redundancy is to understand the real value that Availability, MTBF and MTTR have for our organization. This will give us a reference frame to move on.
Availability is conformed for many issues: Resilient LAN/WAN links, Server Load Balancing, Multiple Internet Links and ISPs, Clustered Servers, RAID, Off-Site Data Backup Storage, Power Protection and Backup, etc.
I suggest to you start looking for information about server clustering. Once you have this issue partially digested you will see that resilient links (internals and externals) and power backup and protection are going to come to the table easily.
You could start finding some info about clustering at Microsoft, Novell, IBM, etc., no matter if the platform is not like yours (some times IT managers need to start evaluating platforms from scratch to meet business requirements and budget limitations):
Other very important issue is to make failure impact analysis on the network components, and isolate the single points of failure.
If you need more detailed information about planning for availability you will find useful the following documents about common mode failures in redundant systems:
http://www.crhc.uiuc.edu/FTCS-29/pdfs/smitra.pdf
http://www-crc.stanford.edu/crc_papers/mitravts01b.pdf
Small organizations are not my specialty. But in the near dozen cases I have worked on as infrastructure consultant the final solution was, for at least half of them, migrating to Novell. This results in more availability and easier administration, while maintaining the investment low enough.
Jose P Mir
jpm@jpmir.net
