RAID Questions

[Richard Meeker]

Speaking of RAID controllers, is there a recommendation for any particular
IDE RAID controller that generally works well with Linux?

>> -----Original Message-----
>> From: Lucas Peet [mailto:lpeet at eccod.com]
>>
>> I have no experience with RAIDs.  The only thing I know is that RAID
>> stands for "Redundant Array of Inexpensive Disks", and that a
>> RAID 0 basically joins
>> all disks in to one 'virtual disk'.
>>
>> > > It has two RAID arrays being used, one is a software array
>> > > using Six Quantum
>> > > 6 gb UW (20 Mb Transfer rate) SCSI drives and I have them
>> in a single
>> > > striped RAID 0, to be used for backup purposes. The second
>>
>> This would be 6*6= 36Gb of storage, right?  My question is:
> Yes, able to use all 36gb.
>
>>
>> > > RAID is a RAID
>> > > 0/1 using 4 60 gb IDE Maxtor ATA100 Drives and the onboard
>> > > Highpoint HP370A
>>
>> ### > controller (which gives me a 104 usable ext3 formatted
>> partition).###
>>
>> What does he mean by this?
> He should have 120gb of usable disk (N/2)*S. Where N = number of disks
> and S = size of
> -=smallest=- disk. I'm not sure why he's getting 104.
>
>>
>> > > > I want to implement RAID Level 1 (mirror) for the /boot and
>> > > / partitions
>> > > and RAID level 5
>> > > > for the balance of my partitions.
>>
>> Also, what do the other RAID levels mean?
>>
> Well there are others on the list much more qualified to answer this
> than I.
> But I have no fear of making myself look ____. So here goes.
>
> RAID 0 is basically taking multiple disks look like one and uses a
> striping
> mechanism to achieve some performance. There is no redundancy here for
> recovering
> crashed drives. Not very useful, there are many better solutions. Disk
> failure all data lost?
> 2 drive minimum.
>
> RAID 1 is mirrored drives. If one crashes no data is lost. This cost
> speed and performance,
> because each virtual drive is mirrored exactly. Highest overhead. 100%
> redundancy. 2 drive minimum. N-1 drives can fail and still be able to
> recover all data.
>
> RAID 0+1 (or 0/1) This is RAID one but with RAID 0 striping on the
> segments.
> This gives redundancy on a RAID 5 level, almost. Limited scalability.
> One disk failure and
> array becomes a RAID 0 system until faulty disk replaced. Two disk
> failure, all data lost.
> Excellent for fileservers. I have seen conflicting stories on this. One
> says 4 disk minimum, another says 2. I think 4 is correct.
>
> RAID 4 two or more drives are used to store data, 1 to store parity.
> Parity disk can be used to
> recreate failed disk. Not sure what happens if parity disk fails.
> Requires 3 disks or more.
> Complex implementation and recovery. Two drives fail all data lost.
> Parity disk becomes bottleneck.
>
> RAID 5 Like RAID 4 but data and parity is distributed among each disk.
> Most difficult to recover from failed disk. One disk failure is ok. Two
> disk failure all data lost. Requires 3 disks.
>
> There are also RAID levels 2,3,6,7,10 and 53. RAID 7 is done only by
> CSC they own it. I have never seen these others, although I hear RAID 6
> is the most fault tolerant allowing multiple crashes and still being
> recoverable. Also very expensive to implement.
>
> That's the way I understand RAID.
>
> I'm sure someone will correct my mistakes,
> Brian
>
>
>