Hardware raid is a hell of a lot faster as long as you are talking about a real raid controller.
Hello, I'm new here. I'm also curious, which would be faster, hardware raid or mdadm software raid?
Am I correct in assuming that mdadm puts more strain on your CPU when writing/reading files?
I think it would be cool if Phoronix wold do some benchmarks![]()
Hardware raid is a hell of a lot faster as long as you are talking about a real raid controller.
I don't have any direct experience, but this is what I've heard: with a small array, under low load, with a modern x86 CPU (typical desktop use), softraid is usually faster; that is, a given request will probably get the data to its destination in less time. Of course, this is at the cost of CPU utilization. As you start looking at performance under load and with larger arrays (e.g. server situations), hardware RAID can start pulling ahead because it's not bottlenecked by contention for the CPU.
Anyway, I think the main reason that "real" RAID cards fetch so much money is that they offer features like hot-swapping and online array rebuilding in a way that allows replacement of a dead drive with zero downtime.
Yeah, it's not as much about speed as to data reliability. If an OS hangs, it'll affect software RAID since it's run on CPU but should be fine with a separate RAID controller.
Software raid can NEVER be *faster* (it is at best AS fast for stupid raid types, like striping) than hardware raid unless those who implemented the hardware raid were a bunch of rabid chimps on acid. Any kind of parity-raid (raid 5, 6, etc.) will ALWAYS be slower in software raid than in hardware raid, especially when the host CPU is under load.
Hardware raid does NOT give you any hotswap and/or online array rebuilding that doesn't ALSO exist within mdraid. Yes, mdraid is happy to allow you to add/remove volumes from the array and/or rebuild the array entirely online with no downtime just like hardware raid.
The advantage of hardware raid is performance, plain and simple. The advantage of software raid is portability (i.e., you can plug your disks into ANY controller and access your array rather than being stuck buying a new card if the card burns out), and cost (i.e., you don't need to buy a raid card for software raid).
FAKERAID (which is a hardware device that pretends to be hardware raid when the raid function is actually entirely within software) is complete crap. It not fast, portable, OR cheap.
Well fakeraid is enough for Win to get pure speed (i use 3x200 GB raid 0 for that purpose) for the boot drive. When you try Linux on it then you can try lots of things to boot from it. From a seperate boot drive you can easyly access dmraid partitions via /dev/mapper/xxx however. My 3.16 ghz quad core has got not problems with some extra cpu usage![]()
Why? All hardware RAID cards I've looked at are doing a substantial amount of their work in software on a general-purpose microprocessor that's a good deal slower than modern desktop CPUs (which makes sense if you think about the engineering constraints). Of course the implementation is thoroughly optimized for running a RAID array, but I don't think it follows that this must be faster than software RAID in all situations.
I guess I could have worded that part better. My point is that if you get a hardware RAID card, those features are advertised and supported by the hardware vendor. You can of course get a cheap hotswap-capable controller card and run mdraid on that, but realistically it's probably going to be untested and unsupported unless you're buying it in a preassembled NAS or something like that.Hardware raid does NOT give you any hotswap and/or online array rebuilding that doesn't ALSO exist within mdraid. Yes, mdraid is happy to allow you to add/remove volumes from the array and/or rebuild the array entirely online with no downtime just like hardware raid.