I'm not sure I buy this as a good idea quite yet. Flash is, right now, about 8x cheaper than DRAM per storage and 10x more expensive than disks. And the DRAM latency is, of course, about 1000x better still.
Basically for many workloads, it's now (edited: original said "not" -- oops) possible to fit entire data sets in RAM, backed by cheaper traditional disks that essentially just stream write logs.
And for workloads where storage costs actually dominate (I'm thinking youtube/dropbox type stuff) it seems like access is mostly streaming, where HD performance is actually quite good.
Obviously SSDs make a ton of sense in the consumer space, where you only get one drive. But if you could pick your RAM/Flash/Disk mix yourself and tailor it to your app, I guess I don't see a lot of advantage to putting your money into the "middle" of that stack.
At scale, disk throughput more than anything else will dominate. How long do you want backups to take? If ssds are faster they might mean fewer servers with more disk throughput per server can give users a snappier experience.
Especially since the article suggests that most of the SSDs are being used for meta-data / content-management with the actual data being on spinning disks, possibly in another data center.
Since they have so much lower power demands I wonder if they are going to move further up the pipe? Instead of the ISP routing a request to an index server which then decides which is the nearest DC or akaimai host with the actual data - would the first line of dropbox/amazon/google/netflix actually move into the ISP's local rack.
The economic crossover for HDD, SDD and distributed main-memory database depends on data volume vs. query volume.
SSD and DMRD win when the query volume is high compared to the data volume since with HDD you'll need to build multiple copies of the system (or equivalent) to handle the load, whereas SSD and DMRD serve more queries with less hardware.
On the other hand, in cases where query volume is low, HDD is cheaper. But if time-to-market matters, you really do want to reduce the turnaround time for your batch jobs and exploratory queries because in the development of KB's with a subjective element you'll probably need to throw 10-100 things at the wall to get the one that sticks.
I can absolutely see Dropbox, Google and Facebook creating a class of SSDful machines as a middle performance layer between RAMful and Diskful machines, but price point and storage limits have a long way to go before there's any wholesale replacement of diskful machines with SSDs.
Also, I'm really surprised there's a datacenter that has a meaningful amount of Facebook, Mozilla, Dropbox and Google servers in the same DC. Looks like someone might be trying to bolster their DC image a bit.
re: bolster their DC image -- equinix doesn't need to bolster anything. $166 a share: EQIX on Nasdaq. They're fairly well renowned for their facilities around the world.
I'm sorry, but share price has almost no meaning without knowing the number of shares outstanding; and then, it only let's you know that the market currently values them at ~$8B. It still seems like a non sequitur, as they are still competing with Raxspace among others. So why wouldn't they need extra press attention?
How much do persistent HD shortages (especially for higher-volume, high-speed drives) since the Japanese earthquake/tsunami and Thailand flooding have to do with moving relative price points of SSD vs. rotational disk?
Also: for app sever storage (and temporary database tables), SSD makes a lot of sense, and contemporary needs vs. available size (100-200 GB is plenty for many system installs, if not a factor-of-ten overkill) makes SSD vs. rotational a no-brainer.
I don't think it has had a much of an impact as I personally expected it to have, though no-doubt it has helped SSDs at least a little.
The biggest effect I've seen is that a bunch of companies (generally big companies, but not HDD creators) have made a shit-load of profit from reselling stock they had before the floods because they realised it would make them more money than using the drives for the purpose they were bought for.
It appears to me that with hard drives the Thailand flood and Japanese earthquake are causing us to lose a generation. Thus instead of going from 1TB mobile drives to 2TB drives, we're having another year of them selling 1TB drives.
This will make flash more competitive as it cuts price per GB in half or by a third over that same period.... but I think it is a temporary effect.
Innovation in spinning rust has not slowed down in the labs, I presume, so I think that when the manufacturing sector is back up to speed they'll be pushing out the same innovations in 2013 they would have otherwise, and possibly skip a generation.
Which is not to say that the writing isn't still on the wall for spinning rust... eventually Flash will catch up in price/GB, though volumes will explode even when its still twice as expensive as the demand will ramp up dramatically.
I have installed 3 in the past month as the prices have come down. I should have done it sooner but did not realize how much they speed up a computer to the point where it is hard to use a computer with a old fashioned hard drive now.
Interesting that Dropbox is making heavy use of SSDs, as I would think that most of their content is more static long-term than similar data at Facebook, for example.
Do flash drives still have a ceiling for how many times you can re-write a particular cell? This is fairly common lore and the article didn't mention it.
Yes, but there are two types of flash cell you can use. SLC cells have an order of magnitude higher write-cycle lifetime than MLC cells, which is why there's a difference in "enterprise" vs. consumer ssds. In practice for most uses an SLC based flash drive will be retired before it reaches its durability limits.
Eh, I just had an X-25M G2 start to fail the other day, after two years of desktop use. Intel replaced it within the week for free, but it was hardly retired.
The x-25m uses MLC flash cells. All other things being equal you'd expect an SLC based SSD to have a 10x longer lifetime. Also, can you confirm that your device failed due to write cycle exhaustion or could it have been due to a firmware bug instead?
Oh, sorry, I didn't realize you were talking about SLC. Yes, that's true. It was cycle exhaustion, as bad sectors started appearing and SMART reported a high end-to-end error count. I had the latest firmware, nothing helped.
Failure rates for flash drives seem to be driven by controller problems or bugs more than actually wearing out the flash. In order to avoid wearing out the flash they implemented very sophisticated software in the controller, and those controllers get wedged as a result (ironically.)
It's crazy how fast the prices are dropping. I feel like they dropped by 40% in the last 6 months. The cheapest 2.5" drive available on newegg is now a flash drive. It's probably hard to make a spindle drive any cheaper than $50 for laptops. I have a feeling the prices are going to similar to usb thumb drives soon.
It's under $1 per GB on SSDs now, which is definitely affordable for average consumers (around $90 for 128GB).
I bought a 256GB SSD in Feb this year for around $360. That exact same drive is now $260.
The beauty is that SSD/Flash pricing is directly affected by Moore's law... so as has happened with RAM, prices halve / capacity doubles roughly every 18 months.
AFAICT hard drives seem to increase capacity fairly linearly, so we might very well have a cross-over within a decade.
When I read this I thought they really meant flash in a general way like USB sticks, heh, not in an underlying technology way. I've read of several people doing such a thing but it seems to be a unnecessary hack anymore. (see http://analogbit.com/node/4 )
Flash is really great in theory, but in my experience with intel and sandforce controllers, it isn't reliable enough. I've bought 2 flash drives and had 4 failures already.
I've seen both drive controllers get themselves wedged to the point where they just shut down and go dead. This happens because they aren't doing a straight 1-1 mapping and along with wear leveling (which is necessary) they're trying to compress (sandforce) and optimize (intel) the way data is stored but their algorithms are too complex to be fully debugged (as of last year, anyway.)
I've gone back to spinning rust as a result. I'd love to move my servers to flash, but not until they go back to really straightforward controllers (I hear the samsung controllers are very reliable, and this is why Apple chose them for their laptops) and don't do anything more than wear leveling.
For performance in servers, I'm going with a lot of RAM.
Right now the drive failure rates of SSDs are way too high in my experience, and reading between the lines and user reviews on retailer sites it seems that my experience is not unusual.
Anecdotal evidence is anecdotal I know, but I've had no issues with my laptop SSD, an Intel 320 Series iirc. At first I was really worried, which drove me to backup properly for the first time, but almost a year in I've had no issues with it at all.
Just curious (and I haven't really used any SSDs yet) if your drive failure experience is with consumer or "Enterprise" SSDs? I'd expect that the enterprise versions would be more resilient but I don't know.
They key aspects for Enterprise drives (both disks and SSDs) are generally 1.) Performance and 2.) When you buy them they come from seperate batches (i.e. a single failure doesn't indicate more are likely to follow).
Servers should basically be created to cope with drive failure in such a way that % of failure isn't an impact on "will this cause problems" but merely on working out the cost of using them.
Actually, the key difference with "enterprise" SSD is that they overprovision the drive to have a cell reserve. That's why you can buy a 512Go consumer drive or a 480Go enterprise drive. As cells get wonky, they are 'replaced' from the reserve.
Purely as an anecdotal data point, I'm a developer so due to constant compiling and other side-effects of my position, I suspect I work my SSD's harder on the write side than average. Using consumer grade SSD's I haven't had any more failures than I did with platters.
They're getting better and better, really fast. Still not quite reliable enough that I've started suggesting to family and friends that they move to SSDs, and if I had to pay for them I'm not sure I would have bothered moving yet (though I haven't personally had a problem in quite some time now), but they're getting there, and quickly.
Basically for many workloads, it's now (edited: original said "not" -- oops) possible to fit entire data sets in RAM, backed by cheaper traditional disks that essentially just stream write logs.
And for workloads where storage costs actually dominate (I'm thinking youtube/dropbox type stuff) it seems like access is mostly streaming, where HD performance is actually quite good.
Obviously SSDs make a ton of sense in the consumer space, where you only get one drive. But if you could pick your RAM/Flash/Disk mix yourself and tailor it to your app, I guess I don't see a lot of advantage to putting your money into the "middle" of that stack.