Unsolved
This post is more than 5 years old
5 Posts
0
1000
ScaleIO sds_network_test_results vs iperf
Hi
I'm getting pretty low performance from my query_sds_network_test_results (my best result seen):
SDS 192.168.1.114-ESX returned information on 2 SDSs
SDS 5d139c4700000000 192.168.30.202 bandwidth 382.1 MB (391259 KB) per-second
Whereas Iperf worst score I've given of tons of attempts: (often its as high as 8.5Gbits/sec)
[ ID] Interval Transfer Bandwidth
[ 4] 0.00-60.00 sec 46.4 GBytes 6.64 Gbits/sec sender
[ 4] 0.00-60.00 sec 46.4 GBytes 6.64 Gbits/sec receiver
The SDS test seems to be presenting the same result I see when using a VMWare Datastore on top of my scaleIO volume, which is a set of 1 single SSD's in each host, running ioperf on it. Which is this:
1) much better reallife measurements than otherhardware connected to the same Backend, because in reallife, the bottleneck is the storage: in this case the SSD is much faster than the 25Drive HDD array.
2) slighly worse read performance, because scaleIO is hitting network limit of between 250-380MBytes/sec, so I see around 270MBytes/sec in ioperf
3) much worse write performance, because the writes are being copied to other storage, so it ends up being about 50% of the write performance. I know the SSD can do more, and the SDS network test result is clearly showing some issue, but iperf running of the SVM itself gets 825MBytes/sec
Any suggestions?