On 08/07/2013 12:55 PM, Maxim Khitrov wrote: > On Wed, Aug 7, 2013 at 10:31 AM, Martin Schröder <[email protected]> wrote: >> 2013/8/7 Maxim Khitrov <[email protected]>: >>> I've read the "Network Tuning and Performance Guide" @ calomel.org, >> >> Ignore that site and search the list archives. > > Understood :) > > I found a number of recommendations for the things to keep an eye on, > but nothing that gave me any ideas on what else to try for improving > the performance.
Several things come to mind: a) is the BIOS current on the test systems? b) power savings options should be turned off in BIOS. 8k interrupts should not be saturating a CPU line that, in my humble opinion. I maintain firewalls running on older Xeons, which serve 4-5 10GbE cards, and when the system is loaded, it can handle 25k or so interrupts before things tip over. c) is the PCIE slot set to PCIe 3 and 8 or more lanes? This is probably a red herring, but some cases "helpfully" reduce to PCIE 1.x speeds unless forced. d) You may want to disable or reduce multiple cores in the BIOS. My testing with multiple cores for each physical CPU actually showed reduced performance and increased variance between performance test runs. e) Keep in mind, if these devices forward packets, both the inbound and outbound interface will be under load. -- -- John Jasen ([email protected]) -- No one will sorrow for me when I die, because those who would -- are dead already. -- Lan Mandragoran, The Wheel of Time, New Spring
