Slashdot Mirror
SATA vs ATA?
An anonymous reader asks: "I have a client that needs a server with quite a bit of storage, reasonable level of reliability and redundancy and all for as cheap as possible. In other words they need a server with a RAID array using a number or large hard drives. Since SCSI is still more expensive than ATA (or SATA), I'm looking to using either an ATA or a SATA RAID controller from Promise Technologies. While I had initially was planning on using SATA drives, I have read some material recently to make me rethink that decision and stick with ATA drives.
What kind of experiences (good and bad) have people had with SATA drives as compared to ATA drives, especially in a server type environment?"
Promise and Highpoint (and any other cheap raid card) in my experience are no more than an IDE card with RAID software that eats up CPU cycles. Recovery options for a lost drive member are usually limited and unreliable. If you want reasonable reliability, go with one of the drives that uses SCSI hardware adapted to an SATA interface (such as WD Raptor). I would personally recommend Adaptec for your host controller needs, as they do the RAID in hardware.
Nice idea, but poor implementation, they have had a tendency to easily come loose on several servers we have.
if you're looking for reliability, this seems like a no-brainer to me. sata all the way. im not aware of an ata drive that even comes close to the 5 year warranty of wd's sata drives.
Gyrate Dot Org - "Where high-tech meets low-life"
I bought a machine from with a controller from Promise and I think I know how they got the name. They kept promising me things.
I was using SuSE 8.2 and they had no drivers but they "promised" that they would be out by the end of the month. Of course I could compile them myself but since that required installing the OS which was impossible without the drivers that required finding another machine and dealing with other problems.
After about 3 months of "promise" after "promise" (this month for sure) they told me it the drivers would be out "in a couple months". The longer I waited the longer away the drivers were scheduled.
It wasn't like I had grabbed 8.2 when it was released either. Promise's Linux "support" was way behind and they basically told me that Linux is their poor stepchild that gets leftover resources when Windows stuff is done.
I contacted my vendor and had them swap the Promise card for a 3-ware. I tossed in the disk and loaded SuSE without any need for downloading or compiling drivers. I'm running RAID-5 on 4 120GB drives. I had a drive fail a couple months back but just hot-swapped/rebuilt it with no problem. The machine was up for about a year before I had to shut it down to replace a failed tape drive but I've had no trouble with the 3-ware.
~~~~~~~
"You are not remembered for doing what is expected of you." - Atul Chitnis
Apple's XServer RAID solution seems to be one of the cheapest dollar per gigabyte solution that I've ever seen. They use fast ATA drives. Although ATA drives can have problems, Apple uses only the best drives from each lot (hence they are a bit more expensive than if you bought the disks from a jobber). The RAID is a true hardware raid, allowing the creation of a hot-spare, e-mail notification, etc. The configuration software is java and runs on any platform. The RAID unit itself is fibre channel, so it can hook to servers running any OS and looks just like a big scsi disk. We have our arrays set up such that we're mirroring 2 phyiscally sparate arrays together (each raid 5+ hot spare), so we can lose up to 4 disks wihout any loss of data. Each array is about 2 or 3 raw terrabytes.
I would avoid the other controller cards you mentioned for the reasons the other posters mentioned. The Xserve RAID is all the benifits of a good scsi backplane (RAID, monitoring, etc) for a fraction of the cost.
Absolutely, we have a 3TB server I recently set up at work with a 3Ware 9500-12 SATA RAID card. The card is expensive (~$700), but well worth it, for the supported RAID levels, management software, drivers, and support that only 3Ware currently offer in this market.
--That's the point of being root, you can do anything you want, even if it's stupid.
StoreCase Technologies
RAID boxen with ATA on the inside, SCSI and/or FC on the outside. Seemingly incredable warrenties of as long as 7 years.
Losing data on an ata raid array happened to a friend of mine and I wouldn't advise using something other than SCSI without understanding the ramifications.
Best regards,
Doc
I made a new years resolution to give up sigs...so far so good!
I have a client that needs a server.
On a related note, I was having dinner at a restaurant and my waiter asked me for a recommendation for a good email program. So I guess it turns outs that I have a server that needs a client.
Whether its just maxtor in general or a few poorly constructed hard drives i've had a few problems with the connectors - simply the plastic tabs at the back had a bad habit of being extremely easy to break :( (i.e to hold cable in place)
You may be right about building a system as reliable, and it'd certainly be hard to compete with it from a size standpoint, you are totally wrong about it being inexpensive.
Apple's 3.5TB system costs $10,999 US. If you were to build a system that comprised 9 Hitachi 7200RPM 400GB drives, you would acheive 100GB more storage space for 3,600$ plus the cost of the server it was hosted in. Throw in 750$ for a high-end RAID card and 1000$ for a server to enclose and handle it, and you're still priced at under HALF the price of Apple's solution.
So, in conclusion, Apple's solution is many things, and is certainly VERY sexy and attractive. But inexpensive compared to a self-built solution it is NOT.
The connection technology with the drive / spindle quality.
(P)ATA and SATA are connection technologies.
They have their individual benefits and drawbacks
(cost, reliability, speed)
The real factors to consider are the details of the drives themselves - vibration dampening, bearing and motor quality, MTBF.
It used to be rather simple to guess what quality of drive you were buying. If it was 146GB or less (73GB, 36GB), and rotational speed was 10K or 15K, it was either SCSI or FC, and an "enterprise" class drive, rated in Mean Time Between Failure.
Good drive, high quality, expect it to last several years, spinning 24 hours a day, sustaining high read and write activity during production and backup hours.
If the drive was larger (200GB+) and slower (7200 RPM), typically an ATA drive, maybe low end SCSI.
Then it was, at best, a workstation class drive, rated in "Contact Start Stops", meaning how many spin-ups and shutdowns the drive should survive. Not meant to run 24 hours a day, and run under heavy load except for short periods.
The lines are beginning to blur with 300 - 500 GB drives with FC drive attachment. Those drives are meant for archiving and reference data. Not production databases and such.
In my personal experience, the 3Ware products are worth the premium.
Pick your attachment technology as appropriate.
Best of Luck,
Patrick (slineyp at hotmail dot com)
Striving to achieve a lower state of conciousness
Performance: Naked drive with linux raid #1 Megaraid/3ware - both slower
I don't know why but how come linux with naked drives using software raid *always* comes in the top with performance. May be you guys can tell me.
- People who believe other people have no right to live, got no right to live ...
The backplanes on server cases are horrid for SATA. They work, but you have to have special hookups for the LEDs (drive fail and activity) and often the controller cards or motherboards don't supply them. All I've managed to get is power LEDs on the front of the Super Micro cases I've worked with.
SATA is not that much faster in practice than PATA, because the kinds of load that you put a drive under in a production environment are not like the speed/load tests used to generate benchmark numbers.
You asked for opinions, and mine is that PATA (ATA-133) is more than fast enough, and the cost of SATA and the quirks that have yet to be ironed out are not worth it. It's the latest shiny object, and shiny objects are not always the most useful.
I base my experience on the Western Digital SATA (mostly 36 gig) drives and the Western Digital 40 and 80 gig JB drives connected to multiple brands of motherboards and add-on controller cards.
I have a client that needs a server with quite a bit of storage, reasonable level of reliability and redundancy and all for as cheap as possible.
So what you need is this.
The ______ Agenda
Here are some of the highlights from their page:
Online capacity expansion and online array level migration
Split mirroring, array hiding, controller spanning, distributed sparing
All RAID levels including RAID5/50, RAID1n/10n
Serial ATA-based
Choice of 4 or 8 channels and 2 functionality levels
64-bit, 133 MHz PCI-X controller in a low-profile, 2U module
And the HIGH-END board can be had for under $350!
(stolen from DaBum) I am dyslexia of borg - your ass will be laminated.
Trying to remedy this by turning off write-back caching severly impacts the performance of the drives and some vendors do not certify the recovery of drives that deactivate write-back caching so this may increase failure rates.
I don't buy this argument one bit.
I agree with you that write-back can break journalling FS guarantees.
However, I don't know of any consumer drive vendor that guarantees that their write-back algorithms are in-order. This means that write-back can trash *any* filesystem, and whether it be RAID or not be damned.
Write-back should *never* be on on drives using modern filesystems.
As for an impact on performance, I call foul again. The write-back cache benefits are useful only in the presence of an OS that does poor disk caching. Take a nice Linux box -- it'll use all available free memory as a big fat writeback cache. There is only a single advantage to using a drive's native writeback controller -- the drive knows the true geometry of the disk (not whatever fantasy geometry it's handed off to the host), and furthermore knows the performance characteristics (settle time, seek times, etc) of the drive. That's useful, but it's not comparable to having ten times or more the amount of memory for buffering.
Hard drive vendors would be *much* better off from a performance standpoint exporting a profile of their drive's performance characteristics to the host -- "settle time on the drive can be determined by this function, seek time can be determined by this function, this is the real geometry", etc. Then, the much more powerful (in both memory, CPU, and code size) host could do whatever scheduling it wanted to try out.
May we never see th
Apple uses our stock firmware. They tweak some constants, but usually everyone who knows how to code a specific drive firmware works for the drive manufacturer. Apple is no exception.
We're running several servers with 3ware controllers and SATA drives where I work and while the controllers are great the SATA connectors suck. They are just too fragile. Everyone in my team who touched the setup - no matter how careful they were - ended up breaking a connector. If you have only one or two cables its alright - but once you end up having 8 or more and try to route them nicely you'll be in trouble.
If you're going for more than just 2 or 3 drives and want to go SATA you should go with one of the newer multilane connectors. One connector carries 4 SATA channels and for an array with 12 drives you only have to worry about 3 cables. That makes the cable layout much neater and the connectors are fairly solid.
No, the PATA cards suffer from being PATA, performance is cut in half due to using the same channel, and the channel is limited to 100/133MB/sec transfer rate. Shouldn't be half, though, unless drives are fast enough to max out the interface speed (which they aren't, fastest 15K rpm SCSI drives are about 109MB/sec internal transfer).
I haven't experienced any issues like that, and can confirm that the hot-swap and hot-spare capabilities work as expected on the 9500-12. I have not performed any benchmarks on the system, but have not experienced any read/write delays, which is probably helped by the cache of the controller and drives. I have 12 250GB drives, a mirrored pair for the OS, and a 9 drive level 5 array, with one hot spare. Pull any drive, and the hot spare works correctly, and the missing drive is rebuilt when plugged back in. Works exactly like the Adaptec 7902/2010ZCR solution in our SCSI servers, except the drives are about half the speed.
The few complaints I've seen about the 3ware cards are very similar to what you've said, though, but seem to be limited to a few specific OS/software combinations, which leads me to believe that: a) Few people realize what impact specific RAID levels have on performance (e.g. RAID 5 requires reads from all disks during a write to calculate the parity information.), and how certain hardware may reduce that impact (or in the case of PATA, significantly increase the impact of that specific problem), and b) There is some sort of issue with a specific type of read being performed by NFS, either a filesystem driver problem, or a hardware driver problem, possibly both.
As a quick summary, the major problem with PATA cards is that they use the PATA interface, command set, and drives. For speed, each drive should have it's own channel. The benchmarks you gave suggest other problems as well, and the older (PCI/33) controllers may have bottlenecks at the bus, but the newer cards (PCI-X/133) eliminate that issue for PATA/SATA drives (since you shouldn't use either if speed is your primary concern). If you want CHEAP, LARGE, and RELIABLE storage, the 9500 works well. If you want FAST and RELIABLE, a dual-channel U320 Adaptec RAID card works wonders, and it's nice to hit 640MB/sec instantaneous transfer speeds, with actual continuous reads at almost 600MB/sec, and writes at about 280MB/sec (RAID 1+0, 6 15Krpm drives, mirrored across channels, striped within channels).
--That's the point of being root, you can do anything you want, even if it's stupid.