Performance Tests using a Rocket RAID 2720 in a HP ProLiant N40L MicroServer
By: John Stutsman
The purpose of this paper is to measure a performance baseline of up to 4 similar Hard Disk Drives in a number of RAID configurations using a Rocket RAID 2720 in an HP ProLiant MicroServer.
I first looked at a Rocket RAID 2720 (“RR2720”) in a MicroServer in HP ProLiant N54L G7 MicroServer – First Look and while I found it interesting (and fun) to work with I noted some limitations including not being able to use my add-on PCIe Intel NIC card or my PCIe WD USB 3.0 card. Without the add-on NIC card a MicroServer with the RR2720 had limited utility as a Hyper-V test box because of the desirability to have a 2nd NIC dedicated to VM’s on a Hyper-V test box. Without the USB 3.0 card I lost the ability of fast server backups over the USB 3.0 connection – another detriment in my opinion. The MicroServer has proven to be an excellent platform for me to test some of the possibilities of the RR2720. The RR2720 is a very popular and proven RAID card on many platforms (including the GA-Z77X-UD5H) that has been discussed extensively in the Forums, Podcasts, and in PCDoc’s world & video’s – so I felt the more I could learn about the RR2720 would be time well spent.
As a result of testing I ran in Performance Measurement of the HP ProLiant N40L and N54L G7 MicroServer I ended up moving my production Windows Home Server 2011 to the N54L. Subsequently, I moved the RR2720 to an N40L and reinstalled the OS (Windows Server 2012) used in First Look from Backups made with Windows Server 2012 Backup Role. Before reading the rest of this paper I recommend you read through the First Look paper. I would also suggest reading through the thread RR2720 BIOS Setting Utility V1.5 Issues.
Before settling on the final configuration that I used for the testing and recording of results in this paper I would install the OS on different RAID configurations to experiment with the look and feel of the system as well as the ease of restoring from a Server Backup. After a few hiccups, detailed in the above thread, the restorations were easy as long as I was restoring from a backup that was the same size or smaller than the new volume I was restoring too. I found that a RAID0 of 2 or more HDD’s, even HDD’s with the “plain” performance metrics as the VB0250EAVER, made for a very robust feel with the OS. In total, I did about a dozen OS restorations to different configurations (the thumbnail at the beginning of this paper is a picture from one of those restorations).
For the testing shown in this paper I restored the OS that I had originally installed in First Look to a pair of Crucial M4 256GB SSD’s in RAID0 on the RR2720 – the Crucial M4’s were in the ODD area of the MicroServer mounted in a Bytecc internal 2.5” HDD/SSD Mounting Kit – this is similar to the Bytecc Kit I used to mount 2 Samsung 840 Pro’s in a 3.5” HDD bay – I had pulled the Crucial M4’s from my Hyper-V server so the Bytecc had the Lian-Li PC-K9WX rail mounts on it. Rather than change the mounts to be able to mount the Bytecc in the ODD with 5.25” to 3.5” bracket attachments I just taped the Bytecc to the base of the ODD bay area for the duration of these tests (I was planning on this only being a temporary installation and later returned the Crucial’s back to the PC-K9WX after these tests).
Figure 1 — Instead of using 5.25″ to 3.5″ brackets to mount the Bytecc 2×2.5″ to 3.5″ mounting Kit to the ODD slots I secured the Bytecc with tape.
Figure 2 — Performance of My OS drive made of 2 x Crucial M4 256GB SSD’s in RAID0 attached to Ports 1 & 2 of RR2720 in the HP ProLiant N40L MicroServer
The performance of the RAID’s that I studied on the RR2720 are summarized in Figure 3 below.
Figure 3 — Summary of RAID Performance DATA
I chose to do the majority of my tests using three VB0250EAVER’s which is the stock 250GB HDD that comes with the N40L’s and N54L’s. Additionally I used for my 4th drive a ST3750630AS which is the stock 750GB HDD original OS drive from my EX-487. The ST3750630AS has very similar performance characteristics to the VB0250EAVER and the HD-Tune performance curve for the 1st third (i.e. 250GB) of the ST3750630AS is very close to that of the VB0250EAVER. While not identical I speculated that using the ST3750630AS would give me results close to what I would achieve with an additional VB0250EAVER — if a 4th one had been available.
I did not do any RAID1 tests because a RAID1 should perform basically the same as an individual drive.
When I finished the series of performance baseline tests with the VB0250EAVER’s, that had been my original plan, I was curious – just for fun — to see what the performance of four ST3000DM001’s in a RAID10 would be. Doing a number of RAID0 tests on the ST3000DM001 did not seem to have a practical real world application in my opinion – though my tests with the VB0250EAVER’s implied that the RAID10 performance of the ST3000DM001’s would be close to the performance of two ST3000DM001’s in RAID0. Additionally, I did not do a RAID5 or RAID6 test on the ST3000DM001’s because I did not see them as practical solutions in the future with 3TB and larger drives for my applications and I was short of time for further testing. In my opinion, a good review of RAID strategies using the very large HDD’s can be found in Discussion of RAID, HDD, SDDs and Storage Configuration. Other reference links can be found in the RAID section of MicroServer HardWare Links.
Figure 4 — Performance of a Single Typical VB0250EAVER — This would also be roughly the same Performance of two of these Drives in RAID1
Figure 5 — Performance of 2 x VB0250EAVER’s in RAID 0
Figure 6 — Performance of 3 x VB0250EAVER’s in RAID 0
Figure 7 — Performance of 3 x VB0250EAVER’s in RAID 5
Figure 8 — Performance of the ST3750630AS that was the Stock OS Drive in a EX-487 and 1st 3rd of HD Tune Curve resembles the VB0250EAVER and I will be used as a Proxy for The VB0250EAVER in Subsequent Tests using 4 Drives
Figure 9 — 4 250GB Drives (made of 3 x VB0250EAVER & 1 x ST3750630AS) in RAID 0
Figure 10 — 4 250GB Drives (made of 3 x VB0250EAVER & 1 x ST3750630AS) in RAID 5
Figure 11 — 4 250GB Drives (made of 3 x VB0250EAVER & 1 x ST3750630AS) in RAID 6
Figure 12 — 4 250GB Drives (made of 3 x VB0250EAVER & 1 x ST3750630AS) in RAID 10
Figure 13 — Performance of a Single Typical ST3000DM001 — This would also be roughly the same Performance of two of these Drives in RAID1
Figure 14 — Performance of Four ST3000DM001’s in RAID 10
Because of the limited compatibility with PCIe add-on NIC and USB 3.0 cards I haven’t been willing to recommend the RR2720 in a MicroServer but — with those limitations well in mind — if your limited use case doesn’t need either of those PCIe cards then the RR2720 would be one way to speed up your HDD access times (with RAID 0) or maximize RAID strategies for increased resiliency (with RAID 1 or RAID 10). (Keep in mind that RAID 0 and RAID 1 can also be set up through Windows OS via Disk Manager and in many cases on a System Board.)
A MicroServer is an excellent platform to take a RR2720 and an assortment of small HDD’s and see what can be done in a controlled setting.