Hi All,
Machine A is a solaris 5.10 HP box (oracle DB) with NIC 2 set to full duplex and 1000Mbps. Machine B is a windows server 2008 R2 SP1 HP box with NIC2 set to full duplex and 1000Mbps. Both the machines are in the same subnet connected directly to each other via a CATALYST 6509-E switch. There is no other network device in between.
I am measuring the network performance as a custom .NET 4.0 framework application which runs on the windows box and communicates with the solaris DB box is not giving the performance as expected. When running tests using iperf (windows as client and solaris as server) with all default options the network bandwidth that I get is around 50% of the actual. But when I set the window size in iperf to be the same as on the solaris box then iperf sends data at around 90% of the actual bandwidth. I am not able to understand how changing the window size of the client (windows server) is affecting the throughput.
Defaults:
Windows server - 64KB default window size - Is sending the data
Solaris server - 48KB default window size - Is receiving the data
Transfer Bandwidth
572 MBytes 479 Mbits/sec
Custom:
Windows server - 48KB custom window size - Is sending the data
Solaris server - 48KB default window size - Is receiving the data
Transfer Bandwidth
1.15 GBytes 982 Mbits/sec
TCP Global Parameters
----------------------------------------------
Receive-Side Scaling State : enabled
Chimney Offload State : automatic
NetDMA State : enabled
Direct Cache Acess (DCA) : disabled
Receive Window Auto-Tuning Level : normal
Add-On Congestion Control Provider : ctcp
ECN Capability : disabled
RFC 1323 Timestamps : disabled