Hi Vincent,
Same stability problems with ext3 and software raid-0 in linux
kernel. Always problems.
That is suprising, I do use software RAID in Linux (RAID-5) with Ext4 and it
works without a hiccup. I must admit the system is very lightly loaded. Rebuilt
times are terrible though. Harvard uni. have their Lustre OSD nodes with Linux
software RAID, so I guess if it is stable enough for them...
However according to logical thinking you first say that i should
consider Reiser-FS for doing a lot of low latency reads and now
suddenly must consider NILFS for exactly the same thing?
I'm not following that logics. Can you explain?
Well, NILFS (as any other log-structured file system) is by design optimized
for writes. Random reads from a magnetic disk would be rather slow (that holds
for any file system of course). But with SSD it is a different story. Because
the penalty for rotational latency and seeks disappears reads are fast and
thanks to continuous nature of all writes in log-structured file system the
write performance is excellent. So generaly the log-structured file systems are
very well suited for flash devices. Google some benchmarks with NILFS on SSD.
Personaly on SSDs I wouldn't use anything but NILFS. As soon as it is marked
stable, that is.
> or keep the whole working set in memory (which is usually
> prohibitive in terms of price).
Chaining really happens a lot
there; a core that is doing a read now
has a high chance of doing a read next time also. In fact odds are
good it is a read to a position closeby a previous done read, which
is why i cache it in the RAM with some small
buffer.
Say a gigabyte of RAM or so in total for all caches of all cores
together.
By keeping whole set in memory I ment to avoid IO completely - like in-memory
databases. In common PC server you can use around 192GB max. today. So, even if
you could afford so much RAM your working set wouldn't probably fit in.
Yeah so they say Amazon personnel living in the clouds (computing)
might have plenty of time right now spamming the net...
Let's hope discussion is about NILFS here :)
Yes, sure. My point was just that there are differrent horses for differrent
courses.
Have a nice day
Ales Blaha
--
To unsubscribe from this list: send the line "unsubscribe linux-nilfs" in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
