Hello, drbd 8.4.4 running over a couple (software raid0/internal metadata) of SSD, local raid0 benchmark is 900MB/sec writing 1,2GB/sec reading, kernel is 3.12.14, interconnects with two Qlogic QLE7342 back-to-back in 40gbit infiniband QDR.
Iperf shows 1,2Gb/sec over tcp (mtu 65520) much more with rdma tests. Resync rate is at 900M, but the initial resync was made at only 500Mb/sec. strange. Formatting as ext4, both dd and bonnie++ show an avarage write speed of about 800Mb/sec, maybe 850-880 range. Which is quite good in drbd connected mode. Kernel is tuned in both tcp stacks and io stack, but I cannot understand why the resync rate is "ONLY" 500Mb/sec. Is the internal metadata updating cost that much even on an SSD subsystem? Anyone? regards, -- matteo _______________________________________________ drbd-user mailing list [email protected] http://lists.linbit.com/mailman/listinfo/drbd-user
