High Performance->Disk-> "Highest Latency"-values due to physical device / command latency on an empty (!) 10GBit ISCSI Network
I am currently testing a 10Gbit ISCSI infrastructure based on cisco SG550XG 10G switching hardware. The esx servers are connected via 10G copper cables (SFP-H10GB-CU5M) to the switches, the fujitsu DX100 S3 Storage is connected via fiber-optics (cisco GBIC 10G-SR SFP+ in the switch and Finisar 10Gb SFP+ GBIC in the DX100 S3). There is currently no load on the switches and / or the storage, still I get the following performance metrics. These numbers can be seen on all of the esx servers even though at different time-slots. But the gaps are the same. Any Ideas where this could come from? I can see that the esx is sending periodically commands (commands issued) - and that at exactly this time the physical command latency values rise. Is this special ESX6 behavior or where could I start troubleshooting this?