On Wed, 2004-03-10 at 17:41, Post, Mark K wrote: > tested, compared to JFS or Reiserfs. There appeared to be a nasty > interaction between Reiserfs and VM time slicing that was causing an > extraordinary number of "SIE dropouts" per second, which was chewing up a
Hmm... I *do* know about a bug in ReiserFS with the 2.4.7 kernel where it scheduled some housekeeping stuff on the journal to happen every 50 mS (2.4.17 corrected that to happen every 5 seconds). While and idle Linux guest with the on-demand timer can do with about 50 ticks per minute, this bug made it do 1250. And that made it stay in queue all day. The reason for the large number of SIE instructions reported is that the guest loads an enabled PSW. So you would see a leight CPU usage and high SIE rate (not transactions, because they're so close together). When we complain about SIE intercepts that is normally because some instructions in some situation cannot be handled by SIE. This would show by high CPU usage and high SIE rate. Rob ---------------------------------------------------------------------- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
