Re: OT: question about sizing swap for solaris

On 05/04/2004 02:18:32 AM, zhu chao wrote:
> Hi, Cary:
> The explanation sounds reasonable. But does solaris use this FIFO to
> manage its VM? If not , more swap space should not cause hard pageout to
> occur more easily, right? Is there any unix that used this kind of VM
> management ?

Belady anomaly is a nice spice to a computer science course but, in practice, I know of no such OS. All operating systems that I work with, with possible exception of one (win 2k), use some kind of LRU algorithm. The rationale behind the LRU algorithm is that if the page hasn't been used for quite a while, it is quite unlikely to be used again, in the near future. That is why students of computer engineering are thought of another important technique to make things right: locality of references. The "locality of references principle" states that the execution points that are close to each other in time should also operate on the set of pages that are close to each other in memory. Good way of breaking away from the "locality of references" principle is to use object orientation, where you don't really know where are the methods stored or what routines do they actually call. What is more, because of the name mangling (talking C++ now), things like prof, gprof or pixie are useless with OO stuff. The efficienc of the LRU algorithm is based on the assumption that the vast majority of programs will respect the "locality of reference". By the advent of OO, many operating systems (Linux is an example) are shifting to MFU (most frequently used) algorithms. That is one of the differences between Arcangelli's VM engine and Rick van Rein's one. The other difference is that the Arcangelli's engine is much better suited for the NUME hardware. Oracle did so with their "touch count" policy which,basically, keeps in memory the buffers with the highest touch count, i.e., "the most frequently used" ones. Also, the "two handed clock" algorithm for the page replacement is being modified to become more "MFU" oriented, especially in Slowaris. Now "bflushd" passes do not turn off "referenced" bits, they're counting "touches". So, the "OO" approach comes with the "FU" virtual memory engines.

-- 
Mladen Gogala
Oracle DBA



Note:
This message is for the named person's use only.  It may contain confidential, proprietary or legally privileged information.  No confidentiality or privilege is waived or lost by any mistransmission.  If you receive this message in error, please immediately delete it and all copies of it from your system, destroy any hard copies of it and notify the sender.  You must not, directly or indirectly, use, disclose, distribute, print, or copy any part of this message if you are not the intended recipient. Wang Trading LLC and any of its subsidiaries each reserve the right to monitor all e-mail communications through its networks.
Any views expressed in this message are those of the individual sender, except where the message states otherwise and the sender is authorized to state them to be the views of any such entity.

----------------------------------------------------------------
Please see the official ORACLE-L FAQ: http://www.orafaq.com
----------------------------------------------------------------
To unsubscribe send email to:  oracle-l-request_at_freelists.org
put 'unsubscribe' in the subject line.
--
Archives are at http://www.freelists.org/archives/oracle-l/
FAQ is at http://www.freelists.org/help/fom-serve/cache/1.html
-----------------------------------------------------------------