Ron Johnson wrote:
On Wed, 2005-11-09 at 10:04 -0600, Mike McCarty wrote:
Ron Johnson wrote:
On Wed, 2005-11-09 at 00:24 -0600, Mike McCarty wrote:
Ron Johnson wrote:
On Tue, 2005-11-08 at 12:02 -0600, Mike McCarty wrote:
Tshepang Lekhonkhobe wrote:
As an example, I tried to find a database which could support
client-server architecture, client and server
on separate computers
8 million records
ten byte unique key, ~50 byte records
8500 probes per second, most of which would fail
to find a matching record, so would
plumb the entire depth of the database
automatic fail-over to another server when server goes down
fail-over must take place in 10ms or less
database can be updated at rate of 100+ records
per second while actively being probed
Alpha VMS & the Rdb database (purchase by Oracle from DEC back in
1994) can *easily* handle this.
It'll cost you, though
I'm not going to go into details, because that wouldn't be
fair to them. And I'm sure that Rdb is a fine DBMS. But
Rdb was one of the systems I evaluated in about 1998 or so.
1998, huh? It could be done back then. A co-worker was at ATT
Wireless back then and they used Rdb for their billing system.
Lots of inserts, lots of reads.
Needed (and needs) lots of expensive h/w, though.
Buy a mated pair of Tandems. In addition to wondering what 'fail over'
was, you could update the OS on the fly. :-)
Dave
|