Comparing query rates (was Re: 9.2.5 db causes high cpu?)
jim at rfc1035.com
Tue Feb 22 10:11:09 UTC 2005
>>>>> "Stefan" == Stefan Schmidt <zaphodb--bind at zaphods.net> writes:
>> FYI, I just did a crude test on an elderly 300 Mhz Pentium II
>> running BIND 9.3. It handles ~2000 qps -- 20 times what your
>> test found -- when answering for in-core authoritative data.
Stefan> Would that be repetative queries for the same record or set?
Nope. queryperf fed a random list of ~11K names that the server was
authoritative for. Whether this was for the same RRset or not
shouldn't have made any difference. The server still has to decode the
query, lookup the name and format an answer. I know some name servers
(PowerDNS?) maintain a cache of recent answers in wire format and can
simply bit-copy the response without doing a lookup and marshalling
the data for a reply if the same query is repeated. Maybe BIND9 will
do this one day.
Stefan> Would you recommend trying out BIND9.3 instead of 9.2 for
Stefan> a recursor or are the changes there, as i supposed, just
Stefan> adding new features like DNSSEC? I.e. any performance
Stefan> related changes in 9.3 lately?
Running the latest code is usually the best policy. It will have the
old bugs fixed. And it'll no doubt have new bugs. :-) IMO the new
features in 9.3 are well worth the upgrade price on their own.
More information about the bind-workers