DSS FC is very fast.

[Robotbeat]

These are the results with an iSCSI volume (1Gb eth with no jumbo frames):


The random write iops never got higher than 117. This is without jumbo frames. My on-board ethernet card doesn't support jumbo frames, but I have a card that does, so I'll use that next. Even so, this is rediculously low performance. There's probably something I did wrong. I'll try to find it. 117 IOPS sounds like it isn't using any of the DSS's system RAM for caching and is just going straight to the disks.

[Robotbeat]

Also, BTW, that's with using a Block I/O iSCSI volume. I'll try it next with a File I/O volume.

[To-M]

THIS IS OUTSTANDING WORK!!! Just to let you know that the engineers are all looking into your work

They wanted to know if you can test with the Atlanta version.

I think I will propose to post your results on our website and see if we can create a section for best performance tests with our customers and partners using FC, iSCSI and NAS.

Thanks Mass Storage - you guy's ROCK!

[Robotbeat]

I finished a huge battery of tests using different test file sizes for comparison. For very small test file sizes (10MB and 40MB) with 4k random writes, fibre channel outperforms iscsi up to 80 times!

The left side is fibre channel, the right is iscsi.


The left side is fibre channel, the right is iscsi.

I have more results coming tomorrow.

[To-M]

Test setting the iSCSI daemon settings for the Target to the following below. When getting better performance - see if we can replace this map with a current one.

Please go to the console and enter crtl + alt + w then select Tuning options then iSCSI
daemon options then Target options then select Target of choice.

maxRecvDataSegmentLen=262144
MaxBurstLength=16776192
Maxxmitdatasegment=262144
maxoutstandingr2t=8
InitialR2T=No
ImmediateData=Yes

Many are reading this thread so we should keep it shorter, but really good job on the comparison!!

[To-M]

Robotbeat!!

Check this posting on speeds reported with Open-E DSS Benchmark with PIC!
Would be good to check with him. I will let him know of your resutls as well.

Server
DELL 860 with 6/i
RAID0 WD GP 750G X2
Open - DSS Version: 5.0up60.7101.3511 32bit
FC HBA:Qlogic QLA2344 2Gbps PCI-X in target mode


http://forum.open-e.com/showthread.php?t=1319

[lufu]

I finished a huge battery of tests using different test file sizes for comparison. For very small test file sizes (10MB and 40MB) with 4k random writes, fibre channel outperforms iscsi up to 80 times!

Hi Robotbeat,

there must be something wrong with your setup. look at my results:



Our DSS is a dual XEON 3,2 GHz, 8 GB RAM, ARC 1280 - 2GB, using block-IO and 1GB iSCSI connections.
We run 4 1TB SATA drives in Raid 10 configuration, in total 5 volumes with 10 TB.
This is no test setup, so we have some load (21 ESX Servers, 136 VMs, 44 SAP Test Servers) on this DSS - the values may be not exactly reproducable.

As you can see, our old and busy setup beats your FC DSS easily in higher file sizes.
It seems to be more constant compared to your very high values with low file sizes and low values at high file sizes.

We actually have no issues with bad performance using iSCSI and block-IO. Its not fast as hell but it is very acceptable with high load.

Best Regards,
Lutz

[Robotbeat]

Thanks, lufu. It's great to get a comparison point!

Yeah, there probably is something wrong with my iSCSI set up.

Part of the problem is that I am using a software RAID 5 using only 3 drives (2+parity).

Not using jumbo frames, either.

Both sides are also using pentium 4 chips with netburst and only one thread of sqlio (since the initiator side has only one cpu core). Shouldn't be THAT slow, though. I tried tuning the iSCSI target daemon settings, but I haven't got much better results.

Part of my whole point was to see how well FC works when just working from cache, since you could load up a 1U server with 100GB of cache instead of investing in lots of SAS drives.

BTW, I really like the Areca controllers. That's what we've used in our setup.

Here are the 4K random write IO results from a single drive hooked up to a 3ware controller with 512MB of raid cache and 2GB of system cache exported via 4Gb FC:


It makes sense that the smaller test files are so high performance and the larger ones are so much slower, since we're limited to a single sata drive on the backend, so we'd be lucky to get more than 60MB/s sustained.