Highpoint RAID cards..

[jerkyjerk]

From what I see Mac OS X Server 10.5, RAID 5 is still not an option for software raid which would be sufficient for what I want to do. Since it’s a 2nd gen XServe G4 any hardware raid card from apple would be out of the question (and budget since its just used for personal stuff) seeing that the XServe G4 don’t have PCI-e slots. My main criteria are:

Card has to be PCI-X(This is a 2nd gen XServe)
RAID 5 to get the most usable disc space
Super performance isn’t a real issue. Reliablity would be more important.
I’m not looking for boot support data only is ok.

With that said, does anyone have any opinions on the Highpoint RockeRAID cards, or any third party Mac compatible RAID cards, I was looking at [url=http://www.hptmac.com/US/product.php?_index=22]RocketRAID 2224[/url] and would like to see what other people’s experience was with them? flaky drivers? poor performance? quirks? good experience?

Thanks in advance,

jerky

[stepansae]

I have that one running for about 2 years now, and can’t complain. WWW interface is not great, CLI would be nice sometimes, Performance across 4 drives in RAID 5 is not great. I run 2.6 TB RAID 5 array across 8 drives. I had one HDD failiure and recoverd from that without a single problem.

s.

[jerkyjerk]

You said performance isn’t great across 4 drives and you are currently using 8 drives. Using 8 drives do you a see a difference at all between 4 and 8? or is it just a marginal difference?

Thanks for the feedback s

jerky

[dpaton]

I’ve used the 1820A for a while and wasn’t thrilled with it’s performance vs more expensive cards, but for what I paid, I couldn’t complain. Under RAID5, the CPU does all the parity calculation, which can be significant. The 3220 has an onboard IO processor, and though I don’t have one yet, I have played with one, and it’s RAID5 and RAID6 performance is significantly better than the other cards on the class of G4 systems I administer.

I think the 2220 and a few others offer firmware RAID5, which is faster than the software RAID5 I’ve been using, but the big daddy is hardware RAID5 processing, and that’s only available on a few cards for the Mac, with the RR3220 being the cheapest and most accessable I’m aware of.

For what it’s worth, the 2220 with fast disks is pretty speedy:

[url]http://www.barefeats.com/hard53.html[/url]

[stepansae]

[QUOTE][u]Quote by: jerkyjerk[/u][p]You said performance isn’t great across 4 drives and you are currently using 8 drives. Using 8 drives do you a see a difference at all between 4 and 8? or is it just a marginal difference?

Thanks for the feedback s

jerky[/p][/QUOTE]

Huge difference. I don’t exactly remember the measured values before expanding to 8 drives, but was something about 3/5 of performance of 8 dirves now (ending at 310MB/s read with 1/3 of the array size free (tested with simple AJA Kona Test, all 8 drives are WD Raid Edition 2)).

Im using this array for network home directories of about 700 users. I have noticed that Rocket Raid performs satifactory for video aplications (final cut), but multiple read and write situations lack performance. XRaid in my scenario performs better (we have some campuses with Xraid, hence the comparison).

s.

[dpaton]

[QUOTE][u]Quote by: stepansae[/u][p]
Im using this array for network home directories of about 700 users. I have noticed that Rocket Raid performs satifactory for video aplications (final cut), but multiple read and write situations lack performance. XRaid in my scenario performs better (we have some campuses with Xraid, hence the comparison).[/p][/QUOTE]

Point taken. I’ve only ever used the RAIDs in my care for a few users at once, never for hundreds. That said, I firmly believe the bottleneck is in the software handling of the RAID5 parity data calculations, and a card with a dedicated processor for such would be a decent candidate for consideration. I haven’t personally tested any however.

[jerkyjerk]

Excellent. This is pretty encouraging for the use on my miniscule network at home, even using just four disks sounds like a viable option. I’ve traditionally relied on just appleRAID for my last several machines but the amount of data I want to put online far exceeds that capacity of the largest single disks available so RAID 1 isn’t really viable any longer. I don’t want to separate all the data across several mirror sets so I’m looking for at least 3TB of usable space with no more than than several concurrent file sharing users. I’ve been mostly concerned with the driver and hardware reliability since I don’t have any experience with the highpoint’s hardware. I don’t want to load up several TB of data to have it go poof because of crappy software. From the people I’ve talked to so far better performance was generally the most griped about thing thing. Stability issues haven’t come up very often.

dpaton’s comments about the RAID processing got me kind of curious how Hightpoints hardware works. Most modern PC motherboards and SATA cards that say they have RAID support usually do their magic in the driver. That can be good and bad depending on the quality of the driver. I think the driver approach yields better performance compared to OS software RAID but can suffer from poorly written drivers. Where the OS software RAID should in be in theory more mature and stable but a bit slower. I’ve posed the question about where the actually RAID processing goes on with the various RR cards to Highpoint sales. Looking at the specs only the 3220 seems to have a general purpose CPU on it(IIRC some embedded Intel type) but hope for some clarification on that. If I get any interesting info back I’ll post it here.

jerky

[dpaton]

jerky-

For home use the priority list goes something like availability, reliability, size. You first want your data accessable. Then you want it around forever. Finally, you want a lot. Unless you’re doing video or audio production work, or streaming full HD content to your TV, a software RAID5 (or 10, if you have the money) will be plenty I think. 4x 1TB gives you 3TB of RAID5. Should be plenty for almost anyone

There are 3 basic ways to do RAID5 and RAID6 parity calculations:
software
firmware
hardware

The first, software, uses the host CPU to do the math. Generally this is a metric f-ton of XOR statements at a very low level to get the parity bytes, and unfortunately, most CPUs are busy doing other things more important than pile and piles of streaming XORs from a disk controller. This is by far the most common RAID5 implementation in the sub-$1kUS card category.

The second, firmware, does the work on the card using an onboard general purpose processor, or using the (normall present) PCI interface chip, and can actually be worse than host-based parity calculation, if the implementation is poor. This isn’t seen very often, but I’ve heard a few cards do have it (RR2220). Don’t quote me on that though. I have no verification that it exists on any card for Macs.

The final, hardware, is the best option, and until the last few years (relatively speaking) was the only viable way to get RAID5 or RAID6 done. This uses a specifically crafted IO processor (like the Intel unit on the RR3220) to perform all of the parity math, and sometimes offloading other storage related processing as well. It’s fast, and generally is about +35 awesome over any other solution. This is how server grade hardware is designed.

[jerkyjerk]

I got a reply back from a Highpoint person and the response was enlightening and raised another question.

He stated that the RR1820A [b]isn’t [/b]a software based card and referred to the RR3000 series as being an IOP card.
I didn’t know WTF IOP meant so of course I googled a bit looking for answers. Dell had an excellent write up on the [url=http://www.dell.com/content/topics/global.aspx/power/en/ps3q03_dumouchelle?c=us&cs=555&l=en&s=biz]various RAID implementations.[/url]

Here’s the meat of the link:

[b]HARDWARE METHODS[/b]
[b][i]I/O processor (IOP)–based RAID[/i][/b]: This is the most common method, and also typically the most expensive, most richly featured, and highest performing implementation. IOP-based RAID performs RAID functions on the disk using dedicated processors, memory subsystems, and peripheral devices. The on-board IOP executes the RAID engine, handles disk array rebuilding, RAID level migration, and other functions such as error recovery. IOP-based RAID can be either embedded on the server as RAID on Motherboard (ROMB) or placed on a host bus adapter (HBA).
[b][i]I/O controller (IOC)–based RAID:[/i][/b] This category uses the processor and memory resources of the I/O (disk interface) controller to execute the RAID engine. Although it looks similar to IOP-based RAID, IOC-based RAID has limited features and functionality because it shares the limited processor and memory resources of disk I/O controllers. Because of its small footprint and low cost, IOC-based RAID is usually embedded on the server platform.

[b]SOFTWARE METHODS:[/b]
[b][i]Driver-based RAID:[/i][/b] This software implementation is integrated within the driver of a specific disk controller. The driver contains the code to run the RAID engine within the OS environment and the BIOS runs the RAID engine in the pre-OS (boot) environment. Within the OS, this approach looks similar to IOP-based and IOC-based RAID in that actual physical drives constituting the RAID volume are invisible to the OS. However, driver-based RAID depends completely on the resources of the system processor and memory for RAID execution, and affects system performance in high CPU utilization environments.
[b][i]OS-based RAID:[/i][/b] This type of software RAID, usually implemented as a filter driver bundled within the OS, uses the processor and memory of the host system to execute the RAID engine. OS-based RAID is independent of disk controller type and uses disk controller drivers for disk I/O.

With that said my question for the Highpoint was to confirm if the all the RR family aside from the RR3000 card(s) are all IOC based designs. Since the IOC based designs are just using the common processor and memory resources of the card to do it’s normal work along an needed RAID processing that would explain why dpaton’s saw a huge diff in performance between the 2220 and the 1820. The cards processors are constantly upgraded like our desktops over time. The newer chips are just faster and as a side affect process the RAID task more efficiently.

jeff

[dpaton]

Jeff-

I believe that Highpoint’s line is broken down like this:
1xxx series: Host based parity (software RAID5), unknown management function for RAID 0 and 1, but probably firmware.
2xxx series: IOP based parity (firmware RAID5), unknown management function for RAID 0 and 1, but probably hardware.
3xxx series: IOC based parity (hardware RAID5), hardwar ebased RAID 0 and 1.

I’m not positive, but after a weekend of reading between the lines on their datasheets, I’m fairly confident of my assessment. In the server I’m scrounging parts for now, I’m probably going with a RR2220, because the price difference between that ($250) and the 3220 ($450) pays for another drive or two in the array.

That said, if Highpoint tells me that the 3220 really does support the 33MHz 64 bit PCI slots in the Gigabit G4 I’m turning into a backup server, I may change my mind and plunk down the extra money for the 3220 and make it a RAID6 box. The old G4 isn’t a fast machine under Leopard, and I need all the hardware help I can get for drives.

[Author]

Posts