On Sat, May 17, 2003 at 01:01:39PM -0700, Ian Clarke wrote: > -----BEGIN PGP SIGNED MESSAGE----- > Hash: SHA1 > > Using VMStat is a cop-out. If we can't find a correlation between any > of the intuitive measurable factors (such as threads active or network > load) then it suggests that something is wrong - and we should figure > out why.
Any such correlation will vary widely between platforms. Kaffe's public key crypto is much faster than Sun's, threading costs vary per OS and per JVM, and you are not addressing the basic problem that there will always exist nodes that cannot handle certain loads that do not exceed their available bandwidth. > > What are we going to do with this load measurement anyway? We will > reduce the number of concurrent requests - yet the whole rationale for > this is that the number of concurrent requests doesn't directly affect > load - so the whole rationale for this falls apart. Less concurrent requests causes less load. More concurrent requests causes better node and network performance (if not overloaded). There is a balancing act, and we have absolutely no idea which side of the pole we are on unless we have something measurable. > > The solution to this problem is to do profiling to understand why some > things affect the load more than others, and fortunately some people > have already started work in this area. That is the solution to a different problem. No matter how fast we make Fred, it is likely to always be possible to overwhelm some nodes without exceeding their bandwidth. > > Ian. > > - -- > Ian Clarke ian at locut.us > Coordinator, The Freenet Project http://freenetproject.org/ > Founder, Locutus http://locut.us/ > Personal Homepage http://locut.us/ian/ > -----BEGIN PGP SIGNATURE----- > Version: GnuPG v1.2.1 (GNU/Linux) > Comment: Debian :: The Universal Operating System > > iD8DBQE+xpUjQtgxRWSmsqwRAuU5AJ9U/cFOTtqo/FTygZib95RQBmxxAwCfVjzG > vf8Zhi4ZEho4iebkOjBws7U= > =AaKn > -----END PGP SIGNATURE----- -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: application/pgp-signature Size: 189 bytes Desc: not available URL: <https://emu.freenetproject.org/pipermail/devl/attachments/20030517/eae09286/attachment.pgp>
