Large database platforms

My group at work is anticipating a large database system requirement. We currently use Sun equipment but I feel that another platform might be needed for this particular requirement. Here are some of the particulars that these folks are anticipating:

1. Database purpose if for network traffic statistics. Traffic data from each of 5,000 routers, concentrators, and other network devices corporate-wide will be stored in the database on a daily basis.
2. Total database size is 6 Gigabytes. This does not include log files, journals, software, etc. Total disk space requirements I feel would be close to 8 gigabytes.
3. There are about 10 different tables; most of them are 1,000 records or smaller but there are two with over 1,000,000 records each. Records are of numeric data and are not larger then 100 bytes each.
4. User access to retrieve data will be done via home grown software which uses RPC's. The interface to the SQL will all be on the server. All user access and data transfer will be over the network. We have an ethernet using OTP and Synoptics concentrators.
5. We anticipate up to 15 simultaneous users accessing up to 2 Megabytes each at any given time. The users wish to retrieve the data in less than 30 seconds when the system is at its busiest.

I have concluded that Sun will not be able to handle this. Even with a 690 MP with full memory, there is still the single I/O processor bottleneck.

Whan I am wondering from you out there is what experiences have you had with machines such as Sequent, Pyramid, Allient, and so fourth with a database of a similar size as this one? Our company as standardized on Oracle so I am stuck with that database unless I can prove that another one can outperform oracle with a significent margin. Besides, I understand (and I hope you can correct me if I am wrong) that Oracle is suitable for a multiprocessor environment in that it has a separate process for each user. f any of you do have such a beast running, I am curious at how much memory you have.

Thanks for your help!

Mark Allyn
allyn_at_netcom.com

 4. They anticipate up to 15 users will need to retrieve data at any

    one time. Received on Mon Jul 06 1992 - 03:43:40 CEST