Hi,
does somebody had done performance measurements of minifilter
vs. legacy drivers (after a legacy to minifilter source code rewrite)
in real-life and/or lab test scenarios? could you share some results,
like % improvements?
it is possible to get less performance with minifilters?
have a nice day,
Sandor LUKACS
Virus Analyst, SOFTWIN
While anectdotal, we have a dual implementation (mini-filter/legacy
filter) combination and we did remark in one specific benchmark scenario
(boot time testing) that the mini-filter implementation was slower than
the legacy filter implementation (95% of the code base is common between
the two) which I did find a bit surprising.
I had to dig up the case I was remembering: the test was “what is the
increase in boot time of two systems, one running Windows 2000 SP4+URP
and one running windows XP SP2.” The answer was: “with the legacy
filter, checked drivers, boot time took 5% longer and with the
mini-filter, checked drivers, boot time took 20% longer.” The 5% time
was actually inside the measured margin-of-error (e.g., we observed
variations in boot time that were +/- at least 5% without any changes to
the system at all.)
Whether you can extrapolate from that is rather questionable and making
comparisons in general is challenging.
Tony
Tony Mason
Consulting Partner
OSR Open Systems Resources, Inc.
http://www.osr.com