7

I'm working at a small TV company. They generates 11TB of video files per month.

What can I do?. I just need ordinary SATA disks, in RAID. What solution should I buy?

(They are using a lot of Iomega NAS)

6
  • 7
    Could you clarify...do you mean they're storing the files, you're looking to expand storage, you're looking to back it up, long term archive...? Are you having trouble with storage space alone, access time,...? What are your needs you're looking to have addressed? Feb 5, 2011 at 23:21
  • I ask because normally the site doesn't give individual purchasing advice for a specific product, but rather will help with general ideas of how to store or save things. Feb 5, 2011 at 23:22
  • Hi. Long Term Storage. I need a new solution, such as dell, expandable ad infinitum.
    – Galleta
    Feb 5, 2011 at 23:27
  • 1
    Don't forget that you'll probably need some type of backup storage, too. That is, unless you don't mind losing all your data when the building burns down. Do you really need all the files to be online at once, or can you go get them from tape when you need them? Feb 6, 2011 at 0:38
  • I'd like to have such a challenge. ;-)
    – poige
    Feb 6, 2011 at 8:16

6 Answers 6

12

Hmmm

"Expanded ad-infinitum" scares me. Do you really need every second of film online there and then? Can you consider tiered storage? Fast storage for recent/highly used stuff; slower higher capacity storage to stuff you need handy, but not snaps fingers right there; and lastly tape or something similar to archive old data?

As for actual options, what do your budgets look like? I'm going to suggest stuff from the Dell range because that's what I'm most familiar with, but all the top tier server vendors will have their own products at the same points but I'm not as familiar with those and you need advice, not a long list of drive models.

I like Dell's Equallogic range of SAN devices if you want to publish this data through more than one servers, and they come in a number of options that can offer you a fast storage tier and a slower higher capacity tier and of course you can buy a tape library of any size you like from Dell too for the archival tier.

If you only want to publish this stuff through one server then consider the direct attached storage option - banks of disks that can plug into a server. Again from the Dell range I know about you'd be talking about the PowerVault range

As I said earlier, I'm not suggesting you have to buy Dell, I don't work for them and I'm not shilling for them - it's just that we use a lot of their stuff and I know it quite well. You can look at HP, Sun/Oracle, IBM and they'll all have similar ranges of products at the same levels, and there are plenty of companies out there who specialise in storage too. But letting file usage grow without managing it like this is a recipe for disaster imho.

I know you're probably being told that they need everything all instantly available, and that you need to do this cheaply, I understand the pressures. But trying to manage data like that without a management strategy is asking for the moon on a stick and it just won't work.

3
  • +1 - you can build either a massive "slow" array by using SATA disks with the Equallogic PS-series (or any other brand's equivalent) or a relitavely large, "stupidly-fast" array using 15k and SSD disks (I say "slow" because in context it is, but 16 disks will actually be fairly fast nomatter which way you mix them) Feb 6, 2011 at 0:18
  • Mark - that's how we've done it, "fast tier" for VMWare farm and "mid-tier" for file servers. Seems to work quite well.
    – Rob Moir
    Feb 6, 2011 at 0:21
  • +1 - Imposing a storage hierarchy works well if the content is accessed in such a manner that it can be mapped onto that hierarchy. If the business can handle recall speed on little-used content being "slow" then it should work great. Feb 6, 2011 at 0:46
7

I work for a broadcast company and we will generate LOTS of video/audio data per month. Our solution was 2 Hitachi SAN, one faster (FC) and one slower (SATA) with lots of cache (256GB cache). We expect for it to exhaust in 3 years. Yeah, HD video is a bitch.

We have a management system in the works that will deal with all content life-cicle and moving it from faster storage (online) to slower storage (near line) and finally to be archived on tape (there's a lot of metadata involved to quickly and easily find the content on the archive later). We have a (expansible) 1000 LTO 4 tape library for archiving.

That's a thing that you should keep in mind. You really just need fast access/space to a determinate time of your programming. Say, the last month. That will be raw footage coming from news teams, recorded studio programmes, live show recordings, plus external productions (syndicated stuff, ads, etc). That is the content that will be going to the playouts and editing and for that must be fast to access (SAN connected to MACs where the editing is done, for example).

After the stuff is aired, you can (and you should) move it to a collection/archive (don't know the correct term in english), probably on tape. If a determinate piece of news, shows or programming is needed later, re-inject it to the faster storage for use and remove it after all is done.

Hope my personal experience helps.

4

If youre looking for a commercial solution, Isilon is one of the cheaper solutions out there and should suit your needs easily. Their product is a clustering solution. You build a storage cluster out of 3+ nodes. Each node has a bunch of disks in it, and the fault tolerance is customizable to where you can lose a node and/or multiple disks and still be operational. You simply expand by adding a new node or an external disk-pack to the existing nodes. Very simple and easy to work with.

After that I'd look at NetApp as a runner up. Theyre a little bit more expensive, but their products offer more features and performance than Isilon does.

1
  • I want to expand on this. Not only can you lose a node or disk (depending on setup, even multiple nodes), the cluster rebuilds the data on its own so that you don't need to rush a new node or disk in, unlike traditional raid storage, where you require the replacement to rebuild.
    – MaQleod
    Sep 30, 2011 at 17:51
4

I assume that this is video data, which means compression and de-duplication won't gain you anything. I also assume that you've decided the big, bulletproof commercial SAN/NAS vendors are too pricey. If everything needs to be online and on the fast LAN (and not on a cloud service, which I assume you've looked at):

Your most cost-effective option will be something like these 45-bay SuperMicro disk chassis stuffed with 3 TB drives. That's 135 TB raw in 4U and < 1500 watts. You can keep buying similar as needed, adding them to existing file server front-ends as required. RAID 6 via good LSI hardware controllers instead of software if you can afford it - read up on the "raid 5 write hole" and be scared of software parity raid.

As storage density improves, you can migrate data from older storage boxes to newer ones, keeping your server footprint fairly constant in the long term. Use tape or removable SATA disks for offsite storage as that is the only cost-effective way to back up this much data.

You'll be buying a 135 TB unit (with drives) about every six months accounting for RAID and filesystem overhead. You'll also need an additional front-end file server every once in a while too. And powering and cooling them all 24/7, and replacing failed drives, and rolling your own management software. Maybe you can use OpenStack to make your own internal cloud storage on these boxes, or live dangerously and risk it all on some of the open source clustered file systems. You'll likely need two or more full-time bodies managing all this stuff.

The cloud is looking better and better, isn't it? If the data must be local and online, I would seriously look at data reduction... after a month or so, re-encode all of your streams (presumably MPEG-2 broadcast) into VBR H.264 video at a much lower bitrate. X264 with '--crf 23' is close to transparent, and half or less the bitrate of most broadcast MPEG2 streams. When presented with the costs, I suspect management will simply live with data older than X days being offline on removable SATA disks or tape.

1
  • FYI, I priced it out, and that SuperMicro box will be US$14000 packed with 3 TB drives. Assuming you use 9-drive RAID-6 arrays, that works out to US$0.13 per usable (decimal) GB. Powering and cooling the box assuming a normal PUE will cost about $2600/year. It looks cheap compared to the cloud, but recognize that staff time for management costs you about $0.50 per minute. Meaning the hardware is the cheapest possible thing, and you should focus on a solution that reduces the amount of people time you need to spend.
    – rmalayter
    Feb 7, 2011 at 20:21
3

If you've got the time to set something up yourself, the guys at Backblaze have a good writeup on how they build their storage pods: http://blog.backblaze.com/2009/09/01/petabytes-on-a-budget-how-to-build-cheap-cloud-storage/

1
  • 4
    These are great if you've got the expertise to roll-your-own, but builder beware: IO throughput will be rubbish because of all the channel sharing. Feb 6, 2011 at 0:16
2

Check GlusterFS: many-petabyte expandability, low cost per gigabyte, management console, appliance-like storage bricks, configurable replication level, as fast as your network allows, open source, etc. The whole enchilada.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .