[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Growisofs input cache (get away from dd?)

Well at least from an architectural point of view I have great doubts if it is 
a good idea to use O_DIRECT for every application but DBMS.
In most cases it is a good idea to rely on the fact that the one who has 
implemented caching in the operating system knew what he is doing. I also 
think that there was scientific research on this topic so people have really 
thought well about the best solution to implement such things.
The benchmarks which were discussed here seem to me synthetic, too. If you 
benefit from O_DIRECT or not has really to do with the circumstances of 
burning while burning your image.
The only case where O_DIRECT is better is, when you burn an image once which 
is located on a mounted file system (for a long time). You need not to 
benefit if,
* You copy it from a network share (smb://, http://) before burning
* You burn it twice
* You checked the ISO before by playing it with VLC
* You just newly created it

Because it really has much to do with the circumstances, you should more rely 
on the operating system, and I do not think that such cases like described 
above are rare.

However, as I did not do a scientific research if, at the end, you will 
benefit from O_DIRECT in more cases than it leads to more disk I/O, I pledge 
for considering the point from page 13 from

"Be careful in using O_DIRECT

If the application may want to use O_DIRECT but it is not self caching and you 
can imagine a setup with enough RAM to cache all the working set of your 
application then you should at least add a switch to turn off the O_DIRECT 
behaviour, so if someone has that much memory he will be able to take 
advantage of it (remember linux runs on the GS 256GByte boxes too ;).

Adding a switch to turn off O_DIRECT can often be a good idea so we can more 
easily measure how much the buffered IO helps or hurts for a certain 

So I will create a patch which adds such a switch to growisofs and post it to 
the list, so it can be discussed for implementing. Otherwise I would need to 
distribute a patched version of growisofs with turbojet...

Reply to: