AW: [Samba] File Systems - Which one to use?

[Dragan Krnic]

>>Have I misquoted you, John?
>>...........................
>>Have you got some new facts?
>
> Nah, I goofed. ext3 is the poorest, but ReiserFS is
> NOT the fastest and still I use it.
>
> I just installed a new machine as my local server.
> Now running SuSE 8.1 Pro with ReiserFS. MB is MSI
> with Athlon XP1700+, 512MB DDR RAM with WD 80GB
> IDE 7200rpm with 8MB Cache.
>
> I first installed Red Hat 8.0 with ext3 and it was able
> to write to disk faster than 100Mbit/sec ethernet could
> deliver the data. During copying of 34GB data over the 
> wire, memory usage was constant at 335MB. I was using
> rsync (not samba) to copy my data.
>
> I then installed SuSE with ReiserFS. Disk write using
> rsync again was slower than 100Mbit/sec ethernet could
> deliver. Memory usage slowly grew until I was using 100%
> of my 1024MB Swap. In fact I found it was faster when I
> turned swap off!

Just the point I made previously, that unrelated problems
tend to obfuscate the relative performances, especially if
not systematically researched. Using "rsync" must have
something to do with the pathological behaviour of memory
usage. There's something about the way Linux tends to use
up all of the memory available that I find strange compared
to commercial *nices. But to go overboard and claim swap
space can only be caused by an ill-conceived app or an
unexpected interaction with an "improvement" in new kernel.

I personally prefer dump|restore pipeline. It never goes
much below the theoretical throughput capacity (about 11.5
MB/s is what I get) and never any swapping. If I dump to 
"/dev/null" than the throughput is way beyond that,
although I can't pull the precise figure off the top of my
head right now (I'll check it up and report in a future
installment if necessary).

Now, clearly I was talking about ext3, because there is no
dump for reiser and because ACLs and EAs are a late
Addition. I used ext3 until recently. Another reason to
favour ext3 was that Oracle for example doesn't certify
any other system for dynamically growing tablespaces
beyond 2GB. (By the way I got about 3-4 times faster imp of
a 22 GB big Oracle exp with my dual Tyan/K7MP1200, 3 GB ECC
DDR, under RH 7.3, than on a top of the line quad HP-PA RISC
440 MHz under HP-UX 11.0 - but it was not a rigorous
benchmark either, so let's not generalize.)

However, using "tar -b 64" I got the same transfer rates
from reiser to ext3 - the limiting factor is fast ether not
fs. I'll see what I get when I redirect to "/dev/null" and
post it later on.

Now the lack of a decent "dump/restore" for reiser is
another sore fact that might stand in the way of faster
acceptance of this fine fs. I'm all sold on the paradigm
and believe that the authors will make it even better
(read faster) in the coming version 4.