Revo 64 RAID card prices slashed

[Steve]

If you read our review of XFX's driverless SATA RAID controller, thought it rocked (which it does), but couldn't justify the ~£120 price tag, then you may want to take another look now. They're only weighing in at £44.65 inluding VAT! Couple that with our free shipping for community members and you're laughing all the way to the RAID creation menu.

Now, all they need to do is listen to my wish and give me a PCIe version dammit!

[ExceededGoku]

/agree , I may consider getting one of these. As it means I could change my motherboard without losing my RAID setup (i assume anyway...)

[johnny02004976]

damn you Steve , i just told myself no more spending and this comes along

[Steve]

/agree , I may consider getting one of these. As it means I could change my motherboard without losing my RAID setup (i assume anyway...)

Unless you're talking about a RAID setup you already have, then yes, you're right.

[chuckskull]

Oh man, this and 250gig drives selling for £50.

Scan are trying to bankrupt me a I swear.

Can anyone enlighten me on one point though? When you add an empty drive and a full drive it splits the data over both, would I be right in thinking if I added two drives with data on, it wouldn't work?

[Steve B]

suppose it depends on ur raid configuration, if you were running in JBOD then i suppose it would just leave all the data as is, until u copied all the stuff to 1 drive, then changed the configuration. not really sure

[chuckskull]

It's just two normal NTFS drives, no RAID.

[directhex]

It's just two normal NTFS drives, no RAID.

there are about four RAID schemes in common use today.

RAID0 glues two drives together and treats them as a large disk, a bit like SLI with hard disks

RAID1 glues two drives together, and treats them as a single disk with an exact mirror copy

RAID0+1 combines the two techniques, with usually 4n drives

RAID5 stripes data across n drives, giving fairly fast and reliable data, with a total volume size of n-1

the Revo64 is a RAID3 device, which is a little like RAID5, but rather than spreading the data and backup info across all drives, you get data-only drives and a backup-only drive (which can be used for rebuilding any dead drive)

you would rarely if ever try to create a new raid volume from existing data-bearing disks, largely because the only RAID mode which can do it is RAID1

[chuckskull]

there are about four RAID schemes in common use today...

...you would rarely if ever try to create a new raid volume from existing data-bearing disks, largely because the only RAID mode which can do it is RAID1

I know how RAID works I'm asking about a specific feature of this board, which is creating RAID arrays based from one filled drive. I was just wondering wheteher it was clever enough to do it with 2.

[Netvyper]

As someone who's only thought about raid thus far, i do like the idea of the driverless card, I've seen such things with SATA Controllers, and im glad to see RAID following

Im curious though, about the 5 port card, a 5 disk RAID3 array would be equivalent to 0+1, and Parity? or 4 striped disks + 1 parity? or something else entirely?

Do you think there will be more of these 'pure hardware' solutions appearing in the future? with Support for RAID5 & so forth?

I think I will keep an eye out for PCI-E Versions of these cards, as they may well herald my migration to RAID. Thanks for a nice post & informative Review.

(edited out comments about RAID5, did more research )

[Funkstar]

ooooooh! the 5 port is only £52. less that a third of what i was going to pay a couple of months ago.

shame i don't have a PC to put one in now

[directhex]

As someone who's only thought about raid thus far, i do like the idea of the driverless card, I've seen such things with SATA Controllers, and im glad to see RAID following

Im curious though, about the 5 port card, a 5 disk RAID3 array would be equivalent to 0+1, and Parity? or 4 striped disks + 1 parity? or something else entirely?

Do you think there will be more of these 'pure hardware' solutions appearing in the future? with Support for RAID5 & so forth?

I think I will keep an eye out for PCI-E Versions of these cards, as they may well herald my migration to RAID. Thanks for a nice post & informative Review.

(edited out comments about RAID5, did more research )

closest to RAID0 with a parity drive, really

and there's not much call for "driverless" cards at the low end of the market - since everyone runs windows, why bother with an actual hardware raid card?

most proper hardware raid cards (read: £300+ cards) are driverless

[kalniel]

I'm waiting for the PCIe version with inbuilt flash, and even USB ports on the card to plug extra flash into.

That would kick bottom in Vista.

[-ChEM-]

when you start using parities theres a lot more to consider than just raid 0 and 1

if anyone is thinking about doing this because its raid0 without the risk of losing data then they need to think more about what is actually happening with reads and writes... with reading its true you will get essentially the same performance as raid0... so if you have a 3xRAID3 array it will perform like a 2xRAID0... but having the parity is going to give people the confidence to make 5 or 6 drive arrays which they would never do with RAID0 so the reads should be much higher (for large defragged files at least, see the RAID0 vs non-RAID debate)... although with PCI cards you'll be stuck at 133mb/s for all your PCI devices combined so it becomes pretty pointless

BUT for writes, the controller needs to know whats on each drive before it can write the parity which means lots and lots of reads if for example you try running your OS on RAID3/5

i think with a large file there should be some kind of intelligent processing where the parity is calculated in-cache and simply written alongside the rest so you have 200mb data which needs 50mb of parity to be written to back it up in a 5-drive array, thats another 25% by volume to be written which isnt that big a deal as you get a RAID0-style striping benefit again to the writing, so a 5 drive RAID3 array should keep up with a 3 drive RAID0 array

the real problem is with small files where you have only a few kb written each time, you still get the extra writing time to write the parity but more importantly whenever you have data that doesnt fit across the whole stripe (the start and end of every file or the start and end of every piece of a fragmented file) you need to read the rest of that stripe in order to write the parity... so in a 5 drive array, if you write a small temporary file across 5 drives, you dont get a good performance because of the lag of the array, and around the start and end of the file the other 1-4 drives each need to be read, then the parity needs to be written then it can move on to the next file... not very fast

im sure quite a few of these controllers will be coming out in the near future but unlike RAID0 where you can be l33t and get your "super double performance boost" at the expense of only a small bit of real world performance, you will really suffer if you bung your whole system onto a RAID3/5 array

(also as raid 3 has the whole parity separate to the rest of the data, the drive which holds the parity will get caned whenever data is written whereas the other drives should get an equal share... so the parity drive will die a lot quicker than it should and also put out more heat)

[Steve]

BUT for writes, the controller needs to know whats on each drive before it can write the parity which means lots and lots of reads if for example you try running your OS on RAID3/5

Not got time to give an answer in full to your post, but to this bit in particular:

Why does it need to be written to the disk, then read to see what it is? The controller knows what data it's going to put where, so it can calculate the parity while the data is still in its write cache, surely?

[Methanoid]

Not got time to give an answer in full to your post, but to this bit in particular:

Why does it need to be written to the disk, then read to see what it is? The controller knows what data it's going to put where, so it can calculate the parity while the data is still in its write cache, surely?

Would have been good if the review had experimented to prove or disprove this theory. I can see this product getting lots of interest but with comments like that some may hold back.