-   Linux - Hardware (
-   -   best hard drive for dvr that constantly writes to disk? (

Incanus 12-10-2003 04:15 PM

best hard drive for dvr that constantly writes to disk?
This is more of a 'Ask Slashdot' question but since I can never get anything posted on slashdot. I figured I'd ask you guys.

I work for a company that says DVR's for security systems. These machines are writing to disk constantly (at least when there is motion on the cameras, which is near constant). One of our biggest headaches is hard drive failures. Customers get really pissed off when you tell them their hard drive has crashed and they lost a months worth of video.

We use mostly Western Digital Special Edition 180gb ATA100 drives. Those are the ones with the 8mb buffer. BTW we don't really see much performance improvement with that extra buffer, but there is usually very little price difference in the two models, so what the hell. We also have quite a few Maxtor 200gb ATA133 drives.

We've had both models fail after 4 or 5 months in the field. Although not every drive fails after that time perioud, some are still going strong after a year or more.
All of our machines are properly cooled. I don't have temperature specifications but we put plenty of fans in the machines. Always put intake fans directly in front of the drive cage.

On the software side we are using Redhat 8.0. The drives are raid striped. Actually we just recently started using LVM so we can extend partitions if a customer wants more storage. I don't think we've had any drives fail in an LVM setup, but its only been a few months since we started that.

I'm curious if there are maybe better brands out there. In my home machines I've always like WD. Do certain drives boast the ability to be able to write to disk constantly? Should we go SCSI? Don't really want to spend the money on SCSI but it may be cheaper than replacing drives constantly. Actually drives are warrantied for so long (usually 3 years) that we always just send them back. Still its a hassle for us and the customer.
Any ideas?

fireman949 12-10-2003 08:27 PM

Could you configure the RAID as 0+1 (Strip/Mirror), or RAID 5. You may already use one of these but you didn't specify. We use RAID 5 on the AIX box where I work. 7-73GB SCSI drives for 365GB total storage. If one drive fails, the parity drive immediately begins writing to the hotspare (hasn't happened yet but we tested it). SCSI is extremely reliable. One of my IT friends told me of several servers they recently replaced that had SCSI RAID configurations and the drives had been spinning for 5-6 years, but the disadvantage is high $$$.

Incanus 12-22-2003 10:20 AM

We use RAID 1 on the system partitions RAID 0 for the main video partition. It would take twice as many hard drives to mirror the video and we are trying to keep cost down. SCSI drives just arent big enough. Well thats not true they make 180gb SCSI's but have you seen the price of those.
Really what I'm trying to figure out here is such a thing as "High Endurance" ide drives. Something that is better than what your typical home user/gamer would use. These drives that we get back have all kinds of unrecoverable errors after only a few months in the field. I'm sure that if we took these same exact drives and put them in a workstation they would run fine for years. But something about the constant write activity seems to be killing them quickly.
I have been searching around on tomshardware for a good comprehensive hard drive review. Who is better Maxtor, Western Digital, Seagate,Fujitsu? I've read some nasty things about IBM/Hitachi.
Or maybe I'm wrong about the drives. Would it be better to use hardware RAID cards instead of software? I don't see how the RAID setup could be making the drives fail like this.

michaelk 12-22-2003 11:27 AM

I looked at the the reliability specs for WD and Maxtor but they really do not tell you anything. Seagate and Maxtor quotes start/stop cycles >50,000 per min for their ATA drives. On the other hand Seagate lists the Mean Time Between Failure (MTBF) for their SCSI drives as 1.2 Million Hours

I would call see if any of the Manufactures have MTBF numbers for their ATA drives.

IMO you should configure several test units with different drives including SCSI and see how long they run before faillure. Another option would be solid state memory but that would be cost prohibitive.

Incanus 12-22-2003 01:56 PM

Thanks michaelk. I did some googling on MTBF and ATAdrives. I came across

There were several other interesting pages on their FAQ

For anyone following this, here are some of the ones I thought were interesting

This one talks about the different brands

I'm usually pretty careful handling drives, but I will think about this page everytime I bump a drive.

It's too bad SCSI is not really a solution for me. We would have to seriously raise the cost of our product to go this route.

michaelk 12-22-2003 03:03 PM

I know that IBM drop tests dummy drives to measure shock loads. This is to simulate what a typical drive can experience during shipping.

BTW of all the drives that I have bought the only ones that have failed were a Seagate SCSI and an fujistu IDE that failed within the first couple of days it was powered. Of course the ones I've dropped don't count :).

Electro 12-22-2003 10:20 PM

Every hard drive will fail but the question when will it fail. There is tons of software that admins uses to know or predict when a hard drive will crash. All of them are not accurate. Many of you may know about SMART. You can go to the manufacture's site and see if they have a linux program for their hard drives. Then you can run the SMART utiltiy. Next you can have it notify the client or user to back up or get another hard drive soon.

If you want sustain write/read rates use IBM. You can use a compact flash disk or a 2.5 inch drive for the OS.

Customers are very careless about their electronic equipment. If it came from someone or the manufacture, the customer will be happly abuse it during the lifetime. You can put all kinds of shock absorbment material around the hard drives to minimize the abuse the customer will do to it. When you minimize the shock, you can also minimize the noise that it is producing. Customers will just love a quiet DVR unit.

You should advertise backup solutions. You can show the customer or client what storage mediums to use for backing up. Also do not forget about off site backups. They come in handy when something real nasty went wrong.

Incanus 12-23-2003 01:05 AM

We have only recently started using SMART. From what I understand so far its not going to save your drive only tell you that its failing. Which will definitely come in handy if we can correct the problem before the data is lost.


If you want sustain write/read rates use IBM. You can use a compact flash disk or a 2.5 inch drive for the OS.
Im not sure what you mean by this. IBM drives better at write/read?
Im not sure where the advantage in a compact flash OS would be.
We are using microATX desktop cases for most of our machines. So there is no place for a 2.5 drive. Although there is probably some adapter for 3.5 internal bay. Still not sure what the advantage would be. Less heat?

I know all too well about careless customers. We do security for a lot of retail stores. The store managers don't really care what happens to our equiptment. They didnt buy it their owners did. The machines get power cycled way too often, which is probably explains some of our HD problems. However, the power cycles are still less than the average desktop.

Im currently looking into backup solutions. Some sort of storage server in a smaller case to run along side the DVR. Off site is not really an option for most of our customers. We sell 4, 8, and 16 camera units. Each camera records about 1.5 gigs a day. Do the math thats way too much to backup off site without a fat connection.

Electro 12-30-2003 07:03 PM

IBM hard drives throughput is much higher than Western Digital. Newer IBM/hitachi hard drives can sustain an output about 48 megabytes per second or more.

Hard drives gets hot. I think the hard drives in your DVR are around 80 F (26.67 C) to 90 F (32.22 C). It maybe even higher than that. Move up to mid-towers or full-towers so you can add heatsinks or heatsink tubing to cool off the hard drives. For rackmount clients, you can use 4U, 5U or 6U.

What I mean off-site backup. I did not mean internet. Internet is still way to slow even with very, very fat connections. I mean tapes. Tapes can hold about 100 GBs. After your clients finish with a few hand full of tapes, they can send to a backup storage service.

You need to look into server environments because what you are supplying to your clients are a server that captures video. Its like a file server, so you need to use RAID-5 or RAID-5 plus another RAID-5 mirroring the first RAID-5 array. Linux is able to do this with out special hardware. I suggest using hardware RAID controllers.

Cycling the power can induce huge current/voltage peaks so you should provide line conditioners, UPS (in-line versions), and surge protectors to your clients.

tarballedtux 12-30-2003 08:23 PM

Seriously how much trust can you put in an IDE drive when almost all harddrives now have only a 1 year warranty down from 3 years. Even if you get one of the new ones with a 3 year of warranty, it doesn't matter you lost your data. $100 refund from the manufacturer won't make your customers not be pissed. Buying 5 HDs to make a RAID array doesn't seem good. I'd say get 2 SCSI drives of reputable manufacturer with a 5 year warranty(Which to me says "We stand by our quality") and make a RAID 1 array.

Just a thought.


All times are GMT -5. The time now is 09:33 AM.