The SSD Relapse: Understanding and Choosing the Best SSD
by Anand Lal Shimpi on August 30, 2009 12:00 AM EST- Posted in
- Storage
The Cleaning Lady and Write Amplification
Imagine you’re running a cafeteria. This is the real world and your cafeteria has a finite number of plates, say 200 for the entire cafeteria. Your cafeteria is open for dinner and over the course of the night you may serve a total of 1000 people. The number of guests outnumbers the total number of plates 5-to-1, thankfully they don’t all eat at once.
You’ve got a dishwasher who cleans the dirty dishes as the tables are bussed and then puts them in a pile of clean dishes for the servers to use as new diners arrive.
Pretty basic, right? That’s how an SSD works.
Remember the rules: you can read from and write to pages, but you must erase entire blocks at a time. If a block is full of invalid pages (files that have been overwritten at the file system level for example), it must be erased before it can be written to.
All SSDs have a dishwasher of sorts, except instead of cleaning dishes, its job is to clean NAND blocks and prep them for use. The cleaning algorithms don’t really kick in when the drive is new, but put a few days, weeks or months of use on the drive and cleaning will become a regular part of its routine.
Remember this picture?
It (roughly) describes what happens when you go to write a page of data to a block that’s full of both valid and invalid pages.
In actuality the write happens more like this. A new block is allocated, valid data is copied to the new block (including the data you wish to write), the old block is sent for cleaning and emerges completely wiped. The old block is added to the pool of empty blocks. As the controller needs them, blocks are pulled from this pool, used, and the old blocks are recycled in here.
IBM's Zurich Research Laboratory actually made a wonderful diagram of how this works, but it's a bit more complicated than I need it to be for my example here today so I've remade the diagram and simplified it a bit:
The diagram explains what I just outlined above. A write request comes in, a new block is allocated and used then added to the list of used blocks. The blocks with the least amount of valid data (or the most invalid data) are scheduled for garbage collection, cleaned and added to the free block pool.
We can actually see this in action if we look at write latencies:
Average write latencies for writing to an SSD, even with random data, are extremely low. But take a look at the max latencies:
While average latencies are very low, the max latencies are around 350x higher. They are still low compared to a mechanical hard disk, but what's going on to make the max latency so high? All of the cleaning and reorganization I've been talking about. It rarely makes a noticeable impact on performance (hence the ultra low average latencies), but this is an example of happening.
And this is where write amplification comes in.
In the diagram above we see another angle on what happens when a write comes in. A free block is used (when available) for the incoming write. That's not the only write that happens however, eventually you have to perform some garbage collection so you don't run out of free blocks. The block with the most invalid data is selected for cleaning; its data is copied to another block, after which the previous block is erased and added to the free block pool. In the diagram above you'll see the size of our write request on the left, but on the very right you'll see how much data was actually written when you take into account garbage collection. This inequality is called write amplification.
Intel claims very low write amplification on its drives, although over the lifespan of your drive a < 1.1 factor seems highly unlikely
The write amplification factor is the amount of data the SSD controller has to write in relation to the amount of data that the host controller wants to write. A write amplification factor of 1 is perfect, it means you wanted to write 1MB and the SSD’s controller wrote 1MB. A write amplification factor greater than 1 isn't desirable, but an unfortunate fact of life. The higher your write amplification, the quicker your drive will die and the lower its performance will be. Write amplification, bad.
Facebook
Twitter
RSS
295 Comments
View All Comments
GourdFreeMan - Tuesday, September 1, 2009 - link
You would, in fact, be incorrect. I refer you to ANSI/IEEE Std 1084-1986, which defines kilo, mega, etc. as powers of two when used to refer to sizes of computer storage. It was common practice to use such definitons in Computer Science from the 1970s until standards were changed in 1991. As many people reading Anandtech received their formal education during this time period, it is understandable that the usage is still commonplace.Undersea - Monday, August 31, 2009 - link
Where was this article two weeks ago before I bought my OCZ summit? I hope this little article will jump start samsung.Thanks for all the hard work :)
FrancoisD - Monday, August 31, 2009 - link
Hi Anand,Great article, as always. I've been following your site since the beginning and it's still the best one out there today!
I mainly use Mac's these days and was wondering if you knew anything about Apple's plans for TRIM??
Thanks for all the fantastic work, very technical yet easy to understand.
François
Anand Lal Shimpi - Monday, August 31, 2009 - link
Thanks for your support over the years :)No word on Apple's plans for TRIM yet, I am digging though...
Take care,
Anand
Dynotaku - Monday, August 31, 2009 - link
Amazing article as always, now I just need one that shows me how to install just Win 7 and my Steam folder to the SSD and move Program Files and "My Documents" or whatever it's called in Win7 to a mechanical disk.GullLars - Monday, August 31, 2009 - link
A really great article with loads of data.I only have one complaint. The 4kb random read/write tests in IOmeter was done with QD=3, this simulates a really light workload, and does not allow the controllers to make use of the potential of all their flash channels. I've seen intels x25-M scale up to 130-140 MB/s of 4KB random read @ QD=64 (medium load) with AHCI activated. I have not yet tested my Vertex SSDs or Mtron Pro's, but i suspect they also scale well beyond QD=3.
It would also be usefull to compare the different tests in the HDDsuite in PCmark vantage instead of only the total score.
Anand Lal Shimpi - Monday, August 31, 2009 - link
The reason I chose a queue depth of 3 is because that's, on average, what I found when I tried heavily (but realistically) loading some Windows desktop machines. I rarely found a queue depth over 5. The super high QDs are great for enterprise workloads but I don't believe they do a good job at showcasing single user desktop/notebook performance.I agree about the individual HDD suite tests, I was just trying to cut down on the number of graphs everyone had to mow through :)
Take care,
Anand
heulenwolf - Monday, August 31, 2009 - link
Anand,I'd like to add my thanks to the many in the comments. Your articles really do stand out in their completeness and clarity. Well done.
I'm hoping you or someone else in the forums can shed some light on a problem I'm having. I got talked into getting a Dell "Ultraperformance" SSD for my new work system last year. Its a Samsung-branded SLC SSD 64 GB capacity. As your results predict, its really snappy when its first loaded and performance degrades after a few months with the drive ~3/4 full. One thing I haven't seen predicted, though, is that the drives have only lasted 6 months. The first system I received was so unstable without explanation that we convinced Dell to replace the entire machine. Since then, I'm now on my second SSD refurb replacement under warranty. In both SDD failures, the drive worked normally for ~6 months, then performance dropped to 5-10 MB/sec, Vista boot times went up to ~15 minutes, and I paid dearly in time for every single click and keypress. Once everything finally loaded, the system behaved almost normally. Dell's own diagnostics pointed to bad drives, yet, in each case, the bad SSD continued to work just at super slow speeds. I was careful to disable Vista's automatic defrag with every install.
My IT staff has blamestormed first Vista (we're still mostly an XP shop) and now SSDs in general as the culprit. They want me to turn in the SSD and replace it with a magnetic hard drive. So, my question is how to explain this:
A) Am I that 1 in a bazillion case of having gotten a bad system followed by a bad drive followed by another bad drive
B) Is there something about Vista - beyond auto defrag - that accelerates the wear and tear on these drives
C) Is there something about Samsung's early SSD controllers that drops them to a lower speed under certain conditions (e.g. poorly implemented SMART diagnostics)
D) Is my IT department right and all SSDs are evil ;)?
Ardax - Monday, August 31, 2009 - link
Well, first you could point them to this article to point out how bad the Samsung SSDs are. Replace it with an Intel or Indilinx-based drive and you should be fine. Anecdotes so far indicate that people have been beating on them for months.As far as configuring Vista for SSD usage, MS posted in the Engineering Windows 7 Blog about what they're doing for SSDs. [url=http://blogs.msdn.com/e7/archive/2009/05/05/suppor...">http://blogs.msdn.com/e7/archive/2009/0...nd-q-a-f...]Article Link[/url].
The short version of it is this: Disable Defrag, SuperFetch, ReadyBoost, and Application and Boot Prefetching. All these technologies were created to work around the low random read/write performance of traditional HDs and are unnecessary (or unhealthy, in the case of defrag) with SSDs.
heulenwolf - Monday, August 31, 2009 - link
Thanks for the reply, Ardax. Unfortunately, the choice of SSD brand was Dell's. As Anand points out, OEM sales is where Samsung's seems to have a corner on the market. The choices are: Samsung "Ultraperformance" SSD, Samsung not-so-ultraperformance SSD, Magnetic HDD, or void the warranty by getting installing a non-Dell part. I could ask that we buy a non-Dell SSD but since installing it would preclude further warranty support from Dell and all SSDs have become the scapegoat, I doubt my request would be accepted. Additionally, the article doesn't say much about drive reliability which is the fundamental problem in my case.I'll look into the linked recommendations on Win 7 and SSDs. I had already done some research on these features and found the general concensus to be that leaving any of them enabled (with the exception of defrag) should do no harm.