<Tenchusatsu, when you ask AltaVista to search for "Elvis", exactly how much difference is there between 512K and 8MB of L2 cache in the server CPU? When 1,000 corporate users each summon their "to-do" lists at 9:00am, will they still be in cache from yesterday? When salesmen in the field query a 40,000-part inventory, will cache play a role in determining response times?>
I'm not a performance validator, so everything I say here will be pure speculation.
L2 cache doesn't work in the manner you suggest, e.g. L2 cache holding the "to-do" lists of 1,000 corporate users overnight. Rather, if anything holds the "to-do" lists of these users, it will be the 1 GB or so of main memory in the server. That's why each CPU could use a huge L2 cache. When the multiple CPUs in the server do their database transactions, they all operate on an enormous amount of data. This might suggest that the memory accesses are so random that no amount of L2 cache can help, but performance tests show that for Xeon, an increase in the L2 cache size does help to increase performance.
<For all the discussion here about latencies, hit rates, etc. has anyone actually replaced a PII server with an otherwise identical Xeon and measured the difference?>
I would suggest checking www.tpc.org, but I'd figure that if you're looking for a valid comparison, where everything is equal except for the CPU, you might be out of luck. The servers built around Xeon are usually more heavily decked out than servers built around Pentium II.
Tenchusatsu |
