AX4 vs P2000 G3: 15k vs 15k drives

Recently I tested an EMC Clariion AX4-5 array and compared it to the HP P2000 G3 array. You can find that post over here, and after I wrote it I was asked to redo the test after installing 15k drives in the P2000 for a more apples to apple test. This post is the result of that test.

Previously the P2000 was configured with 6 x 146GB 10k RPM drives, and we were seeing the following specs:

Now with 6 x 146GB 15k RPM drives, which are still configured in a RAID5, I was able to get the following performance:

With the 15k drives we see a 21.6% increase in IOps on the Real Life test, and 21.2% increase on the Random 8k 70% Read test. The Max Throughput tests were also very close or higher with the 15k drives…. but enough to beat the EMC ?

Well not exactly… while the 15k drives increased speeds in most categories, the two categories where the P2000 lost to the EMC with its 10k drives still falls slightly behind even with its 15k drives. However the two categories where the P2000 previously beat the EMC with 10k drives, is still the case with the 15k drives. Here are the original EMC test results:

One thing that I believe is still handicapping this comparison is that the EMC has 3.5″ drives, while the HP has 2.5″ drives. I wasn’t sure what the expected performance of a 2.5″ drive compared to a 3.5″ drive was so I did a little Googleing and found some comparisons done by Tom’s Hardware. They compare some Fujitsu drives and find that while a 2.5″ is sometimes faster for throughput, they are rarely/never faster in IOps… which is consistent with the comparison between the HP and EMC.

http://www.tomshardware.com/reviews/enterprise-storage-sas-hdd,2612-6.html

Their test show the 2.5″ drives to be anywhere from 50-80 IOps slower than its 3.5″ counterpart. Also access time is about 1ms slower then a 3.5″. So a little simple math shows that if the 2.5″ drives are about 50-70 IOps slower per drive and we have a 6 drive array, we can probably say that we are going to have a 300-400 IOps handicap…. which is almost the exact amount that the HP is behind the EMC by.

Loading

Share This Post

6 Responses to "AX4 vs P2000 G3: 15k vs 15k drives"

  1. I struggle with figuring out how to come up with numbers for my own san. Can you tell me how you calculate these numbers?

    thanks so much

  2. thanks justin ,two things.

    What capacity planning tool do you recommend. Did you just perfmon for the above examples?

    Seems that your Notify me of followup comments via-email doesn’t work.

    good day

  3. Hi Justin,

    when I was a teched 2010 I had a chat to the HP hardware guys about the P2000i’s of yesteryear and their poor performance. As the conversation proceeded the HP rep mentioned the fact that a well known (maybe only inside HP) CPU bottle neck exists in the iSCSI controllers for this series. While the P2000i was slow it did have some nice performance monitoring tools that would show that HP left out of the P2000 G3’s mgmt console.

    Sounds like a bit of a rant know but I wonder at what point with these 1Gbs controllers are we being held back by the controller CPU’s ability to encapsulate for Ethernet or even maybe the HBA’s which are doing your parity for RAID 5?

    My understanding is that with 2.5″ HDD we should be seeing better performance due to decreased seek time, however that’s not taking into consideration more platters which could introduce seek contention (or whatever the industry want to call it).

    anyway my 2 pence 🙂

  4. I know this is an older thread but i felt that is was worth contributing to – apparently the modern (SAS) 3.5″ drives are now sporting 2.5″ platters – meaning their advantages come over to the 3.5″ form factor.

Post Comment