Thrashing ? too many page faults ? very low CPU utilization

1
Thrashing and Working Set
Thrashing ? too many page faults ? very low
CPU utilization Need to reduce page faults as
much as possible Working set (per process) a
subset of a process pages that the process would
need in order to execute smoothly (no or very few
page faults) over some period of time But we do
not know the future need of the process how
would we be able to construct the working
set? Assume locality. What happened in the
recent past will carry over to the near future.
2
Working-set window (?) how far we look back in
time
time
2 6 1 5 7 7 7 7 5 1 6 2 3 4 1 2 3 4 4 4 3 4 3 4 4
4 1 3 2 3 4 4 4 3 4 4
? 8
? 8
working set 1,2,3,4,5,6
working set 1,2,3,4
There seems to exist some locality in this
reference string (where?)
3
Ideally, we should update the working set after
every page reference
t
2 6 1 5 7 7 7 7 5 1 6 2 3 4 1 2 3 4 4 4 3 4 3 4 4
4 1 3 2 3 4 4 4 3 4 4
WS 1,2,3,4,6
WS 1,2,3,5,6
(5 out, 4 in)
Updating at every reference, however, is too
expensive
4
We could instead do the update periodically (eg.,
once every 5 references, as in the following
which uses a ? of size 5)
Note that this method is an approximation
(why?)
Everytime a page is referenced, its reference
bit is set
2 6 1 5 7 7 7 7 5 1 6 2 3 4 1 2 3 4 4 4 3 4 3 4 4
4 1 3 2 3 4 4 4 3 4 4
Timer interruptall those pages whose reference
bit is set is put/retained in the WS clear all
reference bits