FC performs better than FCoE in all-flash array performance.

A study by Brocade claims that all-flash arrays using FC can provide higher efficiency than FCoE. FCoE makes the interface of the all-flash array so that the all-flash array running FC can be easily integrated with Ethernet. However, researchers at Evaluator Group, funded by Brocade, did an experiment to study the differences in bandwidth and latency between FC and FCoE of 16Gb/s when an all-flash array is connected to a blade server. Their findings are as follows:

A single 16Gb/s FC is more efficient than two 10G Ethernet connections.

When SAN workloads take up more than 80% of resources, FC can provide 2 to 10 times faster performance than FCoE

Compared with FCoE,FC, it can save 20% to 30% of CPU resources.

When the device load is 50%, the response time of FCoE is 50% or more lower than that of FC.

With the increase of workload, the instability of FCoE is much more obvious than that of FC, and the failure rate may be 10 times that of FC.

In the test task, three NL video rendering tasks (average bandwidth 250MB/s per task), 126 virtual database applications, 126 mailbox instances and 560 virtual desktop users were used in the experiment. Two manufacturers of blade servers are used on the server.

Russ Fellows, a researcher at Evaluator Group, said: "for FCoE, four 10-gigabyte Ethernet is used (using Cisco UCS server and 6248UP array extender, the connection between them can run a variety of protocols, including FCoE) to connect the server to the fabric. Then it uses 6248 to connect to the 16Gb/s FC interface all-flash FC array.

"for FC, first use a 16Gbit/s fibre Channel connection between the HP / IBM blade chassis and the Brocade 6510 fibre Channel switch, and the same all-flash array with two 16Gbit/s FC interfaces is connected to the switch."

The final experimental result is that using FC can maximize the performance of the all-flash array.