Again same problem: Bad replication performance. We are losing patience now.

Support doesn't answer after 48 hours (ticket # 1014840)
After we called support today by phone with some new information:

Your case is being moved to the second / third level of support
We will require additional time to investigate the problem.
Thank you for your patience - we will get back to you as soon as possible
We are bored now.

Support never solved ticket # 1002902, which was startetd 23.10.2008
and was closed unsuccessful 16.11.2009 with the comment:

This request was send to developers future task list but there is no ETA yet
Support closed ticket # 1011213 (started 27.8.2009) "successful" at 7.10.2009 although no solution by OPEN-E.


Current problem (ticket # 1014840) won't be solved by support in a reasonable time as well,
because first and second level support probably do not even understand, what the problem is about.


OK, maybe some one here can help:

2 identical machines:

Supermicro PDSME+ Motherboard with 4GB mem, CPU Intel Core2Quad Q6600 2,4 GHz

Slot 1 (PCIe 8x)
3ware 9650SE with 8x ST3500320NS SATA drives in Raid 5 (1 of them hot spare)

Slot 2 (PCI-x)
3C996B-T 1000Base-T

Slot 3 (PCI-x)
3C996B-T1000Base-T

Slot 4 (PCIe 4x)
Intel Pro/1000 PT Dualport Server Adapter

Intel Pro Dualport configurerd as Bond 802.3ad and connected via direct cable to same bond on the other machine.
Bond up and running.


Replication of 3 Volumes goes over the bond.
Maximum replication speed 60MB/s, no matter if only one task is running
or all three replication tasks share the boonds bandwidth (then each with 20 MB/s).
Replication tasks are configured with 100MB/s. As long as one of three is consistent,
this should do no harm.


Tried replication through one of the 3com NICs: even worse, maximum repl speed 13MB/s
Tried jumbo frames, no gain

Disk read speed taken from tests.log hdparm is about 224 MB/s,
disk write speed taken from measuring time while initializing volumes is about 150MB/s

But the worse thing of the replication is, that 60MB/s is NOT CONTINOUS.

Watching a single replication task running, speed "flaps" down to 6 MB/s,
raises again to 60, falls back to 30, goes to 45 reaches 60,
falls down to 6, then somewhere around 15 and so on...

Serious problem, since we have to use ISCSI-Failover,
and as we saw before (and this is the only logical thing inside this boxes),
iscsi speed = repl speed.

Any hints ANYBODY ?

Regards.