VSP 360 · Performance flowBlock storage · 3 arrays · 16 Oct 2026, 16:55–17:55 CEST

89% of the past hour's 47,649 IOPS came from three ESXi hosts sharing datastores on Arrays A and B.

Array B carried 32,648 IOPS, almost all of it on one volume, 00:00:47, at a flat 50/50 read/write mix. Array A carried 14,900 and spiked to 27,854 at 17:25. Array C was idle at about 100. Four Fibre Channel ports, CL1-A and CL2-A on Arrays A and B, carried 98% of the load; iSCSI and NVMe/TCP together carried under 400 IOPS.

47,649IOPS
All three arrays
Hour average, 5-minute samples
89%
From three ESXi hosts
ESX-01, ESX-02 and ESX-03, on shared volumes
98%
Through four FC ports
CL1-A and CL2-A on Array B and Array A
32,161IOPS
One volume, Array B 00:00:47
68% of all I/O; 0.06 ms response time
  1. One 10 TiB volume on Array B drives two thirds of all I/O. 00:00:47 ran between 31,209 and 33,113 IOPS all hour, split exactly 50/50 between reads and writes. That flat, even profile looks like a synthetic load generator, not production work.
  2. Array A's spike to 27,854 IOPS at 17:25 came from one volume, 00:00:01, which jumped from about 1,700 to 18,150 IOPS for two samples and fell back by 17:35. The same three ESXi hosts share it.
  3. Paths are balanced on Array B and slightly skewed on Array A. CL1-A and CL2-A each carry 16,100 IOPS on Array B (within 0.1%), but on Array A CL1-A carries 7,706 against 6,900 on CL2-A (53 : 47).
  4. 727 IOPS (1.5%) cross ports with no host-group volume behind them, mostly on the CL7/CL8 port pairs. These are likely replication or external-storage paths between the arrays; Array C presents volumes to Arrays A and B through CL7-A and CL8-A.
Figure 1 · Where the load comes from

Array to port to host type to host group: a single path, through CL1-A and CL2-A, carries almost everything

Average IOPS over the hour, from each array through its ports to the hosts that use the volumes
Band width is average IOPS, on one scale. Ports below 20 IOPS are combined per array. Host groups that share a volume are shown together, because the array counts I/O per volume, not per host. Host groups outside the top seven are combined into Other host groups. Array totals are the sum of their ports, which can run above the array's own counter (Array C: 243 against 101) because ports also carry traffic between arrays. Hover over a node to trace its flows.
Figure 2 · Over the hour

Array B held flat at 32,600 IOPS; Array A ran above 20,000 for fifteen minutes from 17:20

Array IOPS, 5-minute averages, local time
One shared vertical scale from zero. Array C runs along the floor at about 100 IOPS.
Detail

Ten volumes averaged 100 IOPS or more; eight of them belong to the same three ESXi hosts

Peak is the highest 5-minute average in the hour. Read % is read IOPS as a share of total IOPS.