On Wed, 25 Apr 2001, Kenn Humborg wrote:
> I'll try it, but I'm not convinced. Just pulling across NFS (i.e.
> 100% network-I/O-bound) takes 20secs. Decompression (CPU-bound)
> takes 20secs of CPU time. Doing them together takes 40secs,
> while, in theory, the I/O and CPU could happen in parallel, giving
> a run time of 20secs.
hmm... usually i'd defer to your wisdom kenn, but in this case i can
not see how you think that you can reduce 20secs for reading, 20secs
for gzipping to less than 20+20secs.
you can shift where the 20sec for reading occurs by prereading, but it
has to happen irregardless. you can not make gzip compress data that
hasn't been read yet though.. :)
i'd say fastest option (if sticking to nfs) is just let gzip do it and
avoid extra copies of user data from one process through kernel to
gzip.
> However, using raw TCP will give me a 32K buffer in the kernel,
> which may be enough to stay ahead of gunzip.
>> I'll be back...
>
having invented time travel no doubt.. :)
> Later,
> Kenn
Maintained by the ILUG website team. The aim of Linux.ie is to
support and help commercial and private users of Linux in Ireland. You can
display ILUG news in your own webpages, read backend
information to find out how. Networking services kindly provided by HEAnet, server kindly donated by
Dell. Linux is a trademark of Linus Torvalds,
used with permission. No penguins were harmed in the production or maintenance
of this highly praised website. Looking for the
Indian Linux Users' Group? Try here. If you've read all this and aren't a lawyer: you should be!
