Soft RAID vs. RAID Card

[frimple]

Hello All,

So I thought I would post some numbers to start a discussion and/or fuel the fire(?) on the soft RAID vs. hardware RAID discussion. I've had a few intel drives and a RAID controller for a while but work seems to get in the way of benchmarking/setting up a system. So now I would like to start a thread about the system and the results that I've seen. If course YMMV and these are environment specific results, but here's a little something to start the conversation. Anyways to the numbers...


Environment
------------------------------------------------------------

System
Mac Pro 4,1 (2009)
3.2Ghz W5580 Processor
6 Gb Ram

Soft RAID
2X Intel G1 SSD's (80Gb)
ICH10R onboard controller

Fast RAID
2X Intel G1 SSD's (80Gb)
Areca 1231ML (512Mb cache)


Results
------------------------------------------------------------

Soft RAID
Sequential

Uncached Write (4k)    -- 125.3 MB/sec 
Uncached Write (256k)  -- 59.35 MB/sec 
Uncached Read (4k)     -- 52.40 MB/sec 
Uncached Read (256k)   -- 439.92 MB/sec

Random

Uncached Write (4k)    -- 121.25 MB/sec 
Uncached Write (256k)  -- 48.99 MB/sec 
Uncached Read (4k)     -- 16.49 MB/sec 
Uncached Read (256k)   -- 297.76 MB/sec

Fast RAID (Areca RAID Volume)
Sequential

Uncached Write (4k)    -- [B]501.38 MB/sec[/B] 
Uncached Write (256k)  -- [B]576.83 MB/sec[/B] 
Uncached Read (4k)     -- [B]95.97 MB/sec[/B] 
Uncached Read (256k)   -- [B]646.00 MB/sec[/B]

Random

Uncached Write (4k)    -- [B]188.50 MB/sec[/B] 
Uncached Write (256k)  -- [B]595.72 MB/sec[/B] 
Uncached Read (4k)     -- [B]92.56 MB/sec[/B] 
Uncached Read (256k)   -- [B]647.20 MB/sec[/B]

Interpretation
------------------------------------------------------------

Of course the controller's cache is the main reason for the huge difference. However, I would like this thread to be a resource for anyone who is asking if they need a hardware RAID solution or if Soft RAID will be sufficient. All four of the SSD's in this test are destined for the hardware controller and it's cache will be upgraded to 2GB soon. As soon as I have new results I will post them for everyone to see here.


Notes
------------------------------------------------------------
I used XBench.... the worst benchmarking tool on the planet. However, it was free and I was slightly intoxicated when I got around to doing it so

 

[belvdr]

However, writing over the network is going to drastically reduce the throughput when using the hardware controller. A single gigabit connection can ideally transfer 125MB/sec and only if both sides of the connection can transfer data that fast. Again, that's ideal, but I've seen 100-115MB/sec regularly.

 

[frimple]

However, writing over the network is going to drastically reduce the throughput when using the hardware controller. A single gigabit connection can ideally transfer 125MB/sec and only if both sides of the connection can transfer data that fast. Again, that's ideal, but I've seen 100-115MB/sec regularly.

Certainly in a SAN or NAS situation you're not worried about throughput at this level as your network is your bottleneck. I wanted to get some numbers up so people can see the difference between a Soft RAID solution vs. a dedicated Hardware RAID card at their machine.

 

[VirtualRain]

Can you run some real world testing using a stop watch for say loading a bunch of high-res images into Photoshop, or even boot times, or other tasks like that?

 

[frimple]

Can you run some real world testing using a stop watch for say loading a bunch of high-res images into Photoshop, or even boot times, or other tasks like that?

Boot times are usually pretty poor as the RAID card has to initialize it's firmware. I also don't have photoshop, but I can open all the applications on my computer at once?

I'm planning on getting the other 2 drives installed today and loading Snow Leopard on it, I'll post results when complete. Also if you have any benching utilities you'd like me to try (that are free) let me know. Once I destroy the Soft RAID partition though it will be gone for good.