|
The server should be dedicated to running the chosen network software to maximise the response times.
| |
 |
An example file server, used in a live 4XTRA network, is a Compaq Systempro using a single processor board with an Intel 386 processor and 32 megabytes of memory (as cache to improve disk response). It has twin 220 megabyte hard disks, which automatically act as mirror back-ups of each other; we have had one disk failure, which was fixed by simply removing the offending disk and inserting a spare (cannibalised from the backup machine), no software action was necessary, the new disk was updated automatically by the Systempro while the overnight run continued, without marked degradation! |
| |
|
The NOVELL software is capable of detecting and by-passing minor memory defects and has build in facilities to "roll back" to a fully updated version of a file upon recovery, if there is a sudden power failure or some similar interruption to service. |
A note on Server Contingency
| |
|
In the case of catastrophic failure of the file server, a backup file server, which is identical in all ways, should be kept. |
| |
|
This would not have an up to date copy of the system, but an overnight process can sign on to the backup file server and copy the start of day positions and all other files automatically. |
| |
|
If the file server fails during the day, the positional files can be recreated automatically from the start of day positions and the "today's" deals file by the system recovery processes, which run on the background machines. The deals files are automatically copied by all of the background machines to their local disks and by the monitor station to the backup server. |
| |
|
The other files which are change are: static data files; rates files; and sundry files controlling dealer preferences etc. The rates files are mainly derived from the external rates feed. The static data and sundry files are handled by a direct backup by the monitor station. The change logs should also be copied to allow a manual check of integrity in event of failure. |
| |
|
If the room containing the server and background machines was damaged (for instance by a fire) all this information would be unavailable. The ideal situation is be to place the backup server off site and updated by the monitor station; a small number of PCs attached to this server can provide a contingency site for covering positions in the case of a complete loss of the building containing the dealing room. If the problem is less catastrophic, the backup server can be attached to the dealing room network at any point normally used by a dealer workstation. |
|
