Discussion
Hi,
I have 2 500gb external drives which i have been keeping synced with a software utility up to now. but it's a pain, and i often have to go through manually and fix stuff. So I have decided to probably convert them to RAID 1, using the inbuilt software RAID on Mac OSX. Someone told me software RAID wasn't very good because if there was a crash during a RAID operation then it could mess everything up. Is this true, and is it likely? I have had maybe 3 kernel panics on this computer in 18 months, 2 of which were due to a program which I have since stopped using.
Second question - if I set up the array, can I take one of the drives away, plug it into another computer by itself, and read the files off it as if it wasn't normally part of a RAID array. Obviosuly I wouldn't write anything to it.
Thanks for any help
I have 2 500gb external drives which i have been keeping synced with a software utility up to now. but it's a pain, and i often have to go through manually and fix stuff. So I have decided to probably convert them to RAID 1, using the inbuilt software RAID on Mac OSX. Someone told me software RAID wasn't very good because if there was a crash during a RAID operation then it could mess everything up. Is this true, and is it likely? I have had maybe 3 kernel panics on this computer in 18 months, 2 of which were due to a program which I have since stopped using.
Second question - if I set up the array, can I take one of the drives away, plug it into another computer by itself, and read the files off it as if it wasn't normally part of a RAID array. Obviosuly I wouldn't write anything to it.
Thanks for any help
JamieBeeston said:
Software raid is window dressing, it works, but you wouldnt EVER put it in a live environment... but for home, meh, whats the worst that could happen.
Thanks for that. It's not really just for home - it will be holding a total archive of all our clients photos for the last 2 years, so I really don't want it all dying on me. If the worst does happen, would it actually render the drives unreadable?
Most likely not, but software raid will actually hamper performance, as IDE drives need to use the CPU for I/O control (unlike hardware raid / scsi devices which have their own CPU) this means your processor will get a hammering just from the RAIDing,
if the worst happend, it would probably just boink the one file in question, whereas on a Hardware solution, it would exit 'more gracefully'
if its for work, spend the $$$
if the worst happend, it would probably just boink the one file in question, whereas on a Hardware solution, it would exit 'more gracefully'
if its for work, spend the $$$
dcw@pr said:
these are external firewire 800, does that make a difference?
External drives might make a CPU load difference, in that the CPU usage might actually be even higher.
I dont know for certain, but with USB, quite alot of system resources are used handing the device.
Only way to tell realistically would be to check CPU load as your write to / read from the disk.
Enjoy.
No substitute for a good backup anyway. What happens if the building burns down? No business? or do you plan to pull one of the member volumes on a regular basis to take offsite?
Never had a problem with VMS host based volume shadowing (software mirroring to you folks) but then again, with uptimes measured in years what can you expect.
Never had a problem with VMS host based volume shadowing (software mirroring to you folks) but then again, with uptimes measured in years what can you expect.
TeamD said:
No substitute for a good backup anyway. What happens if the building burns down? No business? or do you plan to pull one of the member volumes on a regular basis to take offsite?
I take one of the drives off site with me - that's why i wanted to know if it would work as an independant disk when i did that.
Okay, well this is the way I see things...
Get at least 3 drives, that way you can always have a reasonable backup offsite if the worst does happen. And cycle them in a manner that you are happy with.
Copy and merge operations will probably take a bit of cpu whilst they sync up but general operation shouldn't be heavily impaired. Bear in mind that with mirroring you should see an improvement in Read performance, although obviously you're doing twice as many writes. What form of access do you anticipate most of? Reads or writes?
In my opinion, you probably have as much chance of the volumes being corrupted due to system crashes as you would if you were writing to two separate drives at the time of failure. But nothing is absolutely bullet proof, it's all a case of what you can live with within your price range, with an unlimited budget anyone can get hardware solutions with battery backed cache dual redundant controllers power supplies etc etc etc
Get at least 3 drives, that way you can always have a reasonable backup offsite if the worst does happen. And cycle them in a manner that you are happy with.
Copy and merge operations will probably take a bit of cpu whilst they sync up but general operation shouldn't be heavily impaired. Bear in mind that with mirroring you should see an improvement in Read performance, although obviously you're doing twice as many writes. What form of access do you anticipate most of? Reads or writes?
In my opinion, you probably have as much chance of the volumes being corrupted due to system crashes as you would if you were writing to two separate drives at the time of failure. But nothing is absolutely bullet proof, it's all a case of what you can live with within your price range, with an unlimited budget anyone can get hardware solutions with battery backed cache dual redundant controllers power supplies etc etc etc
TeamD said:
Bear in mind that with mirroring you should see an improvement in Read performance, although obviously you're doing twice as many writes. What form of access do you anticipate most of? Reads or writes?
on a typical hardware platform, sure, on a CPU intensive platform (no external controller to offload the I/O processing) then you might even see a drastic reduction in reads under heavy system load.
Your idea of syncing is again ok in theory, but even at 40mb/s, thats still 3.5 hours at full pelt to resync 500gb (assuming fully populated and unchanging as it syncs!)
that would be a heavy cpu load..
Raid is great, so long as you give it the hardware it requires.. which is dedicated hardware..
JamieBeeston said:
Your idea of syncing is again ok in theory, but even at 40mb/s, thats still 3.5 hours at full pelt to resync 500gb (assuming fully populated and unchanging as it syncs!)
that would be a heavy cpu load..
I was thinking of a situation where one could allow the sync to complete outside working hours in much the same way as a tape backup. But I do accept that software raid is not an optimal solution, once again. it's all down to readies innit?
JamieBeeston said:
Your idea of syncing is again ok in theory, but even at 40mb/s, thats still 3.5 hours at full pelt to resync 500gb (assuming fully populated and unchanging as it syncs!)
that would be a heavy cpu load..
Having thought about that again is it 40 megabits per second or 40 megabytes? If it's bits then surely it would take 8x as long?
But even worse, remember, a full sync requires (at the very least) 1 read from the source device, a read from the destination device, a comparison and then a potential write to the destination yuk!!!!
But Hell!, 500gig, even a HP GS1280 with 2Gig Fibre SAN attached StorageWorks EVA Virtualising Disk Array would chew on that for a bit! Hmmm...previous test results for a 200gig chunk of storage mirrored by host based volume shadowing (software) full sync = 4hrs 20minish so for 500gig = 10hrs 50mins!!!!!!!!! On industrial strength kit! (8 CPU god knows how many gig memory etc) But obviously, if the shadowing was down to the HSV controller rather than software things would happen somewhat quicker
I admit defeat Jamie, you're right I obviously wasn't thinking earlier, either that or I was distracted by the thread about Strip Joints
>> Edited by TeamD on Wednesday 12th January 17:46
TeamD said:
I admit defeat Jamie, you're right I obviously wasn't thinking earlier, either that or I was distracted by the thread about Strip Joints
You trying to distract me ?
latest firewire operated at 800mbit, but the drives themselves being ide tend to max out around 40mB/sec anyway, hence my calcs
dcw@pr said:
latest firewire operated at 800mbit, but the drives themselves being ide tend to max out around 40mB/sec anyway, hence my calcs
Not if you have decent drives...
I can get 65MB/sec out of mine
A true 65mb/sec out of non scsi
You're having a Giraffe..
maybe from the cache on them
and even if you could, they dont use 'decent' drives on these firewire caddies
be honestly interested to see a HD Tach report showing 65mbsec sustained over a disk tho (ie not burst)
dcw@pr said:aah yes, the larger caddies do have multi drives inside, striped up to appear as one, so that could easily explain it.
Swear to God - sustained transfer. I think that these cases actually contain two drives, maybe they are running some form of RAID 0 inside?
thinking it over, its not outside the bounds of possibility for drive technology to have come along to 60mb/sec i suppose... either way
dcw@pr said:
it's one of these if you're interested
www.lacie.com/products/product.htm?pid=10491
we have 12 of those at the Datacentre. useful little boxes.
They doing upto 1.6Tb iirc
Gassing Station | Computers, Gadgets & Stuff | Top of Page | What's New | My Stuff