Petabytes on a budget

Interestingand certainly not recommended for a production environment. I think the article Daz gives tells you why.

The best solution I've seen i the Boston 24TB server which comes in at £10k, steep but so much better than that hodge bodge server
 
I suppose speed and performance don't really matter here, seeing as how they're used for online backup. I'd have thought the speed restriction would come mostly from the consumer. Still, all the other oversights, especially using desktop drives, are stupid.
 
D D Danneh said:
I suppose speed and performance don't really matter here, seeing as how they're used for online backup. I'd have thought the speed restriction would come mostly from the consumer. Still, all the other oversights, especially using desktop drives, are stupid.
Click to expand...

Using a large amount of desktop drives, in theory isnt too bad. If they are in a decent array of course and how long the server will be in service!
 
Ouch, Not sure I would risk my data on that thing even if it is 10x cheaper than competitors....
 
The thing you need to remember is that due to the way they work, even if one whole machine nuked all of it's hard drives, no data is lost.

Similarly, this is why hotswap isn't a requirement, because other servers are available with the data while they take this one down to swap discs and rebuild arrays.

It's an ingenious storage solution that is perfect for it's task - ie. very high volumes of only slightly redundant data for peanuts.
 
dangerstat said:
Interestingand certainly not recommended for a production environment. I think the article Daz gives tells you why.

The best solution I've seen i the Boston 24TB server which comes in at £10k, steep but so much better than that hodge bodge server
Click to expand...

What the hell, hodge bodge?

I think some have missed the point, those "pods" aren't single solution, each one may contain a copy of the data, of which there could be hundreds of copies. Write speeds may be slow, but then who knows what level of de-duplication is being done and how its all written to. Raw disk bandwidth is only a small fraction of the equation.

It may not be a perfect design, but it's hardly a bodge and I can't believe you think it is.
 
These are amzing designs for one purpose, massive storage, fast loading times for data to the pod and simple as they use a mesh (sort of) topology accross many many pods.
Think of these as building blocks rather than complicated OS hosting email running file servers in offices. This is big boys toys!

Google work on similar principles abeit a huge scale over these but still they work on failed % of kit in a rack before replacing hence they have rack pods.

All about the scaling baby.
 
I wouldn't be surprised if the likes of Amazon S3 ran a similar setup.

Google run desktop grade equipment in there clusters, i wouldn't call there setup a bodge job. ;)

Dell, HP etc simply cant offer a cost effective solution for these requirements. So its easy to see why company's look at in house solutions.
 
visibleman said:
Why not? :confused: It fits the goal for cheap storage where performance/uptime isn't a priority (data is replicated across multiple 'pods').
Click to expand...

Well do they have any geographic redundancy for starters? The article indicates not.
 
You must log in or register to reply here.
Back
Top Bottom