Scale of network statistics

[To-P]

The axes are auto-scalable, they will change the maximum value according to the current transfer rates on your NIC.

[blindzero]

I attach the images I mean.
The auto-scale you mean is what I already know, but the values seem to be strange.
You can see eth1, eth2 and eth3.
Eth1 is Gigabit Cross-Link for replication
Eth2 and Eth3 are Switch-connected Gigabit ports.





But as I wrote: one possibility is that the title is "wrong" and it should be "Byte" not "Bit"?
What especially doesnt make sense to me: the "unlimited" switch-ports have less performance then the cross link connections on which we have set some limit in the replication task configuration.

Thanks,

Matthias

[blindzero]

The build is 6.0up75.8101.5377 64bit

[webguyz]

Are you doing MPIO across ports eth2 & eth3? If so then that would make sense that the eth1 replication link would be double the other two.

[blindzero]

Hi webguyz,

yes we do MPIO on eth2 and 3 and the double make sense for replication, BUT:
- the initial replication is finished
- the replication link on eth1 is a GBit link...so a peak of 80 MBIT (as written in the title) is still much to little

That was the reason for my thread. Because
- if it is really BIT its just to little
- if it is BYTE its quite ok (could be better ;-) but the title is wrong...

Matthias

[blindzero]

Anybody any experiences? Think have to open a case...

[Toni Bizjak]

I've noticed the same.

Volume replication speed is real 80 MB/sec, but "statistic" shows approx. 8 times lower speed 80Mb/sec=10MB/sec.

Or maybe, because SCSI is used, there is no statistic for it. ?!?


Using the newest 6.0up90.8101.5845 64bit
and
Intel 82598EB 10-Gigabit AT2 + upd_0947-DSS-V6.upd