dark_inferno :
yes i realize that it can run raid 5 but a lot of people told me thats its considered a fake raid and does run properely because it uses the main cpu, but i could be wrong
There are a few different ways to run a RAID, all with their advantages and disadvantages:
1. Motherboard-based RAID. This is what you'd get if you went into the BIOS and set up your RAID there. The pros are that it's essentially free and that your OS sees the array as a single volume and can install onto the RAID volume if you provide drivers at boot time (such as the infamous Windows XP RAID floppy.) The cons are that onboard SATA controllers are typically overwhelmed amount of I/O that a 4x 1 TB disk RAID5 can generate while at full roar and bottleneck your RAID's performance (often noticeably so), you can't migrate the array to a newer, different motherboard very easily, and that the ability to even see the array may not be present if you don't use an official supported OS (such as if you want to run Linux or Windows 7.) Your CPU also does all of the RAID parity calculations.
2. "Dumb" discrete-controller RAID. This is an add-in card that contains a disk controller but has the CPU do all of the RAID parity calculations. The pros are that these cards' disk controllers are typically a ton better than motherboard onboard southbridge controllers, particularly the ones on PCIe x4 and x8 cards, the performance is often as good as hardware RAID cards but for a lot less money, and that you can move the array to different machines by moving the card around. The cons are that these cards still use the host CPU for parity processing and that if your card dies, you need to get a similar card to recover your data.
3. Hardware RAID cards. These are add-in cards that contain a disk controller, cache memory, and a little CPU to do parity processing. The pros for these units are that they also have very good disk controllers and that there is almost no load on the main CPU for RAID calculations. You can also move the array to different machines by moving the card. The cons are that these units are
expensive, your data is also tied to that particular kind of card, and that performance is not always a whole ton better than the "dumb" discrete cards.
4. OS-based RAID. The OS sees all of the disks individually at boot and then sets up the RAID during the boot process. The big advantage of this is that it is hardware-agnostic and very easy to migrate to other hardware. You only need to be running the same OS on another machine to be able to see the data in the array. You can also run an array at a partition level, so you can have more than one array per set of disks, such as having half of the disk be a RAID1 and half be a RAID0. The con of this method is that you cannot install the OS to the array itself in most cases, since the array gets set up after OS install. There are some exceptions (you can have the OS on a RAID1 in Linux as you can define the RAID1 after OS install) but generally the OS has to sit on its own non-arrayed partition.
Personally, I like to use a "dumb" discrete controller card and run OS-based RAID. The "dumb" card lets me drive a lot more disk I/O through it than I can most on-motherboard SATA ports, plus I get the advantage of not tying my array to any specific pieces of hardware (motherboard or controller card.) Yes, the host CPU computes parity information, but that is a very small load on any even halfway-modern CPU. I use this method on my seven-year-old dual 2.66 GHz Xeon file server and the parity calculations took up maybe 10% of one core at worst for my three-disk RAID5. I wouldn't waste your money on hardware RAID controllers unless you have huge disk I/O levels and are wanting to run it on puny little single-core CPU like the Atom 230.
Facebook
Twitter
Instagram