HPE Smart Array (Gen8) RAID Card Benchmarks

It was difficult to find a direct comparison between the two Gen8 RAID cards and their performance with SSD storage. I had been experiencing issues with performance on my SSD arrays, as had one of my coworkers who has a similar setup. We both decided to order a P430 and run some tests, and if the results were good, order a bunch more cards to replace all the P420’s in our servers.

Gen8 Smart Array Controller Overview

There are a bunch of different models of Gen8 Smart Array controllers, however, this benchmark focuses on the P4xx series of cards, as these seem to be the most common. It’s worth noting that the H240 and P440 are also reported to use the same raid controller chip as the P430, although there are likely some driver differences between the model years.

Testing Overview

Test Setup:
DL380p Server – Balanced Performance power mode
2x E5-2670v2 CPUs
192GB DDR3

//To eliminate as many variables as possible, these tests did not use the integrated P420i controller in the DL380p server that was chosen for this test. I had a spare P420 addin card, and although they should be identical, I wanted to be sure the results were as accurate as possible.

These tests compare the two different RAID controller types over a variety of tests. This is also to investigate if caching helps or hurts an all-SSD array environment

Caching Tests

All tests were performed on a single WD Blue SSD. Enterprise SSDs typically handle drive cache differently, this is mostly to see how the caching reacts with consumer SSDs.

About Smart Path

HPE has a feature called “SSD Smart Path”. It disables the SSD’s own cache and the controller cache and writes directly to the flash. This might work fine with enterprise SSDs, but with consumer SSDs you will see ~40MB/s writes. Do not use Smart Path.

P420 Controller

Controller Cache Enabled, Drive Cache Enabled

You can see a huge spike at the beginning, then it settles down at ~500MB/s. It seems to be dumping some kind of cache every few seconds, but it doesn’t seem to hurt performance too much.

Controller Cache Disabled, Drive Cache Enabled

For some reason, the performance here is pretty bad. I’m not exactly sure what’s happening, as it’s worse than in a regular desktop, even though this should be as close to the same config as a non-server as possible.

Controller Cache Enabled, Drive Cache Disabled

The performance here is also pretty good. This is showing that the controller cache makes a pretty big difference, and not just for the first ~2GB of data.

Controller Cache Disabled, Drive Cache Disabled

Ok, this one is confusing. The WD Blue 1TB should get a MUCH better speed directly writing to the flash. This must be the controller doing something stupid.

Array Tests

Now that we have determined the best caching setup for SSDs, we will run tests with various array types to see how the controllers handle it.

All tests run with Controller Cache Enabled and Drive Cache Disabled. The drive cache does seem to provide a small performance boost, but at only about 6%, it’s not enough IMO to offset the potential data loss disadvantage.

P420 Controller

2x 80GB P3500 SSD (RAID 1)

Slow, but consistent. These are 80GB drives so this speed is about what I’d expect after the cache has been consumed.

1x 1TB WD Blue SSD (RAID0)

Same as in the caching tests, this result is pretty much in line with what I would expect

2x 1TB WD Blue SSDs (RAID 0)

Ok, WTF happened here. The speed was ok for a bit and then tanked to 3MB/s. The disk activity during this test was at 100% even though it was only writing 3-8MB/s for most of the time.

2x 1TB WD Blue SSDs (RAID 1)

Same as RAID 0, but slightly less bad. It seems the P420 isn’t very good at doing actual RAID.

P430 Controller

2x 80GB P3500 SSD (RAID 1)

1x 1TB WD Blue SSD (RAID0)

Almost identical to the P420

2x 1TB WD Blue SSDs (RAID 0)

2x 1TB WD Blue SSDs (RAID 1)