Unsolved
This post is more than 5 years old
52 Posts
0
498
September 24th, 2013 07:00
Expected performance of srdf/a over GigE
How to troubleshoot GigE link performance
We have 4 RA ports locally and remote within a dedicated replication VLAN. vmax to local switch / router into MPLS WAN and same at remote.
Using symrdf -g groupname -rdfg n query -i 5, we observe values of c. 150 MB/s initially but this quickly drops to 10 - 20 MB/s.
This example was an adaptive copy of 2 x 33 Gb disks.
Are there any vmax based tools we can use to validate link performance - what might be a reasonable expectation for performance over a 1 Gb/s link ?
If anyone aware of any docs on this subject please advise.
No Events found!
RRR
2 Intern
•
5.7K Posts
0
September 25th, 2013 02:00
I've seen speeds up to the theoretical performance of the link. So for a 1Gb link that would be about 100 MBps. Per link, so if you have 2 fabrics, each connecting over 1Gb to another site, you're using 2 x 1Gb links, so 2 x 100 MBps = 200 MBps. If the network is 100% working the way it should work. But I guess it depends on your network then.
What also helps is using as many disks (not LUNs, but rotating disks) as possible on the source and as many as possible on the target. This way you'll avoid bottlenecks of having too few disks. As a rule of thumb a single disk can deliver 10 MBps. What you're seeing might be because at first a lot of disks start copying data to the other site, but in the end only a few disks are doing the work, therefore dropping the speed to the bottlenecks of those few disks.
Because your speed initially starts at 1500 MBps I assume you have a disk bottleneck as described. The RA ports are hardly a bottleneck in your configuration.