Oracle and Red Hat Collaborate [LWN.net]

Oracle and Red Hat Collaborate

Posted Oct 29, 2003 1:59 UTC (Wed) by StevenCole (guest, #3068) [Link]

In Ingo Molnar's post announcing the 4G/4G split:

the typical cost of 4G/4G on typical x86 servers is +3 usecs of syscall latency (this is in addition to the ~1 usec null syscall latency). Depending on the workload this can cause a typical measurable wall-clock overhead from 0% to 30%, for typical application workloads (DB workload, networking workload, etc.). Isolated microbenchmarks can show a bigger slowdown as well - due to the syscall latency increase.
i'd guess that the 4G/4G patch is not worth the overhead for systems with less than 16 GB of RAM (although exceptions might exist, for particularly lowmem-intensive/sensitive workloads). 32 GB RAM systems run into lowmem limitations quite frequently so the 4G/4G patch is quite recommended there, and for 64 GB and larger systems it's a must i think.

So it seems likely that Red Hat will provide other kernels for users who don't need the 4G/4G split, which will probably mean most.

Oracle and Red Hat Collaborate

Posted Oct 29, 2003 17:25 UTC (Wed) by seanegan (subscriber, #15672) [Link]

I can assure any readers that the 4G/4G can be very detrimental to performance. I know of benchmarks of machines running several single threaded poll(2) driven network servers with random access to nearly the whole 4G space. The benchmarks showed a slowdown that made the server nearly useless.

Clearly other usages can have a positive impact from the 4G/4G VM split. I just want to point out an example usage where the 4G/4G VM split can hurt much worse than the advertised %30 max speed slowdown. I imagine that the example I've given is due to the impact of lots of syscalls and cache flushing being worst-case.