IOPS performnce with HP SSD and Raid controler 440ar

Why do you except more write IOPS? In a RAID1 the write speed is determined by the slowest drive in the RAID. There is no write-performance boost. And 10000 IOPS write can be okay, depends on the drives used.

The other thing: How many RAID controllers does your ML350G9 box have? Not that you share a single controller for both RAIDs.

What are you using for benchmarking and how many threads is it running?

We have a logicial drive in RAID1 with the 2 SSD. There is only 1 RAID controler HP 440ar. We need more IOPS  for MS SQL apllication performance.

I do my HDD test with AS SSD Benchmark.

If they are SATA you could re-run the test using the motherboard SATA ports, can't bypass the RAID controller chip if they are SAS (well you can but it's all or nothing so the 5 others would stop working).

What kind of application? When "optimizing" MSSQL applications by throwing hardware at the problem, the first thing is to check the installation and the actual bottleneck.

Here more RAM (read performance) and having the correct number of tembdb files (1 file per core and when contention is a problem add 4 files in each mitigation step).

For testing the RAID I would use DiskSpd.

And to see, how good or bad the numbers are: Test your SSD's not connected to the RAID and then in the RAID1. Post the numbers.

Whether it's intended for SQL or not they should be getting a far higher result, at least with 64 threads. The single thread results from AS SSD Benchmark are pretty meaningless especially for write since that's as much a test of the server's CPU, RAM and bus speed and completely irrelevant unless they intend to restrict SQL to one thread. I hadn't tried it before so I asked how many threads they were running not realising it did single thread plus 64 thread test each time it was run.