We have a (partly) gigabit/copper internal network at the institute, and the MTU can be set to 9000.Maybe someone with experience in fast networks can comment on how average packet loss and latency affect the maximum TCP window you might want to use. With an MTU of 1500 or lower, over the Internet, do your windows actually go beyond 65535 bytes, even if you enable scaling?
The nodes which have to access the Oracle DB only have 100 Mbit links, and deactivating wscale on them had no measureable effect on the network throughput between nodes and internal file servers (which have Gbit links). One reason for this is, I suppose, that this is about _TCP_ window size, whereas NFS (the transfer method we mainly use) is _UDP_ only under Linux, thus unaffected.
If you consider gigabit/copper a fast network and can suggest experiments/meassurements, I'll be happy to conduct them.
Cheers, Steve
_________________________________________________________________
Add photos to your e-mail with MSN 8. Get 2 months FREE*. http://join.msn.com/?page=features/featuredemail
