1 of 1 people found this helpful
Forgive me for not answering your question.
But relax. Don't worry about it. You will not be able to write-wear out (Intel) SSDs with any kind of normal workload before the drive is obsolete anyway. Even the consumer drives are rated for 20GB of writes each day for five years: Intel® Solid-State Drive 320 Series Product Specification
And if it is the repeated writes/overwrites of the same files you are worried about: Don't be. The physical sectors that are presented to the OS are not mapped directly to a specific physical cell on the drive. When you overwrite the same file over and over again, it is not written the same physical address on the drive. The wear leveling takes care of that. There are also a lot of people recommending that you don't store your OS swap file on the SSD. No! The SSD is the perfect place for a swap file. That old recommendation is from back when people were using compactflash memory cards as improvised SSDs. Write wearout is not a problem anymore.
Case in point: I have been using an old second-gen 80GB X25-M for around two years as an OS drive and more. During that time, I have written a total of 3.46TB to it and the wearout indicator it has only just now gone from 100 to 99 (indicating how much spare flash is left). This drive will be utterly obsolete before it is worn out from writes.
You want to have your random writes on the SSD, so you can get the most benefit of the OS/application speed increase it provides compared to spinning storage. Don't worry about writes.
I do not like the "don't worry about it" answers for these kinds of questions. It's like saying, if you've bought a car and it is guaranteed to last 3 years, therefore you don't need to bother doing anything to make it last 7. My workplace still uses 7-year-old PCs. My auntie has had the same PC for 8 years. I've observed some applications writing tens of gigabytes per day to my SSD, particularly for out-of-control verbose log files or download cache folders.
Someone will eventually write a dedicated application to give you a report on the "top talking" applications and files over a 24-hour or 1-week time period, but until that time the best option I've found is to use Process Monitor from Sysinternals. Apply a filter to operations beginning with "Write", if required add another filter rule to limit it to the partitions that are on your SSD, and then enable the option to drop filtered packets.
You can then use the "File Summary" window to see the files that are written to the most, and from that you can drill down using additional filters to see which application is responsible for the writes to each file.
All use of third party applications is entirely at your own risk. As usual, be sure to regularly back up your data.
1 of 1 people found this helpful
I fully respect that stance and your argument. And I commend you for taking the time and effort to do what I strictly failed to do, which was actually answering foler's question. And may I add: Everyone can of course treat their SSDs in any way they feel is right.
So please forgive me for continuing to argue my stance on why I think it is a waste of time to micro-manage OS and applications to keep the writes to the SSD to a minimum.
My main argument is exactly that even if you don't micro-manage the writes, it will still last far longer than is practically needed (not considering other causes for drive failure than write wearout). I am not completely buying the car analogy, in that (if we stay with the cars for a moment) this car will probably not fit in the road width we have by the time the engine wears out. You might as well use it fully while you can drive it on the road. I will present some real-world test data to back up that claim and try to calculate some realistic lifespan predictions.
Intel's (and other's) official write endurance ratings are actually quite conservatively set. How much so? Some great guys at the http://www.xtremesystems.org forums are trying to find out by writing different SSDs to death: http://www.xtremesystems.org/forums/showthread.php?271063-SSD-Write-Endurance-25nm-Vs-34nm
Just to pick a Intel SSD model from a theoretical worst-case capacity/wearout standpoint, let's look at the 40GB 320 Series. It finally gave up after having 685TB written to it (and verified). The s.m.a.r.t. attribute Media Wearout Indicator hit 0 (meaning theoretically worn out) at 190TB so let's use the low number of the two to have some safety margin. And Let's drop it to 170TB just for some extra margin. You mentioned tens of GBs of writes a day as a plausible scenario. Let's say 30GB per day. That's a minimum of (170000GB/30GB)/365 = 15.5 years before it stops working due to write wearout, by which time we will probably have a hard time finding a (working) computer with a SATA connector in it. We can also do the calculation on the actual completely unrestricted average write load I had on my OS/app SSD during a two year timespan, which was around 4.8GB/day: (170000GB/4.8GB)/365 = 97 years before practical wearout. The numbers speak for themselves.
A secondary argument: One of the reasons you buy an SSD should be for speed. By moving (especially random) writes to spinning disks, you take away some of that application/OS speed increase. So you would be suffering less-than optimal speed and a more complex OS/app setup and maintenance for a drive lifespan extension that is arguably pointless.
smarg, thanks for connection to process monitor. This is what I need.