Cache Writing - Enable or not to enable?

[Edster]

Hello,

Could anyone tell me the advantages and disadvantages of disabling cache writing on the priming HDD? Thanks in advance

 

[ignatzatsonic]

I think the default setting is usually to enable it?

I have a dim understand of it from memory. As I recall, enabling it provides slightly faster performance but has a slight risk of data corruption in cases of a bad shutdown--like from a power outage. It can also have consequences when using RAID.

The geekiest thing to do is to enable it, but use a UPS (uninterruptible power supply) to eliminate the risk from power failure.

Me? I've always left it enabled and haven't knowingly suffered any negative consequences.

 

[Edster]

Thanks for your help, I guess the best thing to do then is to keep it enabled. I had read somewhere though, that it helped performance?

 

[ignatzatsonic]

Yeah, it is supposed to help performance. How much, I don't know.

If you disable it, caching is not used, and you don't take the risk of losing or corrupting data on a bad shutdown. I think that means that if disabled. all write operations go directly to the disk (not to the cache). Since they go to the disk directly, you don't have the risk of corruption. If enabled, and you had a bad shutdown, the data sitting in the cache has not yet been written to disk and could be lost. That's my feeble attempt at understanding.

You could easily download HD Tune and do some tests with it off and on to see how much difference it makes. Matter of fact, I might try that myself.

 

[ignatzatsonic]

Here ya go. These results are for my boot drive, a WD Caviar SE16 WD6400AAKS 640GB HD.

The first graph is cache writing enabled. The second graph is cache writing disabled.

Note that the transfer rates are negatively affected when caching is off--more of the blue line is in lower territory. The maximum is the same, but the minimums and averages are different. The burst rate and access times are unaffected.

This is the result of one test only, but should be representative, at least of my drive. A hard drive engineer may well laugh at these tests and maybe correctly so--I don't know how much reliance to put on these results as they apply to the real world.

 

[Edster]

That is extremely helpful and I thank you for your efforts
I might have a play around with it now!

 

[Edster]

I can see your point now. On mine, the enabled writing cache seems to be the fastest on average, I'll keep that enabled from now on.

 

[ignatzatsonic]

Looks like caching helps you very little on your drive. Maybe because the drive is relatively slow by itself and has only a small cache?? You might be a candidate for leaving it off--you might never know the difference and wouldn't have any corruption risk?

 

[jimbo45]

Hi there

In (I hope) plain English

Caching come into its own when you've been using an application for a while (or when the system has learned how you usually use your computer).

What it does is saves often used data into a specially fast memory area (not in RAM but in the disk controllers OWN memory area) so that if you need to access this data the disk contoller can supply it straight from the cache instead of having to create a set of I/O instructions to find the correct sector on Disk, position the drive and do the data transfer - both READ amd WRITE.

On a new computer the system will need to do some form of learning before the cache is effectively used.

Some better algorithms even do a "Data Pre-fetch" which reads data before you actually need it - based on your previous usage of the computer --it does this whilst the computer is idle so it won't degrade the performance at all.

Successful caching algorithms (very tricky to write) can ENORMOUSLY speed up a computer - especially when data bases are being accessed with similar queries.

You'll also find that a disk with a larger cache costs quite a bit more than a similar capacity drive with a smaller or no cache.

(This is why you should also "Quiesce" a USB external drive before unplugging it -- the system will need to write any remaining / pending I/O requests from the cache back to the drive).

Cheers
jimbo

 

[Edster]

So in the long term, it is better to enable caching?

 

[ignatzatsonic]

As I understand it, enabling it helps from a speed standpoint and hurts from a data safety standpoint.

So you get to define "better" however you wish--faster or safer; one or the other; not both.

Most people are obsessed with speed and will choose to enable it, even though if the switch had been in the disabled position for years, they probably wouldn't have ever noticed it or complained. It's mostly about benchmarking and obsessing over minor details.

Looking at your HD tests, it seems to make virtually no difference on your drive.

I don't have a clue about the real-world risks of losing or corrupting data. I have never knowingly had data corrupted by enabling the cache, nor do I use a UPS. I have a power failure perhaps once a year.

 

[pparks1]

Plain and simple...you want write caching enabled for best performance. And the best bet is to put a UPS on your computer that way if you have a power outage you won't lose any data that was committed to cache and not yet written to the hard drive. UPS's are cheap these days. $40 to $80 will be more than enough for home use. I've got 4 of them at home and all of my PC's that I use have one.

I recently ran into a nasty performance issue on a high end server at work which didn't have a battery module on the RAID controller. Without that, VMWare would not write cache anything on it's own and performance was abysmal. Adding the battery to the RAID controller memory meant that anything in memory in the event of a power outage would be maintained by the battery. Once that battery was in place the server controller began write caching and performance shot through the roof. Of course, this example doesn't translate perfectly into the home world market ...but the general idea is that it provides performance.