DNS performance Help when query log is off -- which default parameters will impact the DNS performance

[PENG, JUNAN]

Hi, Tony

During my performance test,  I also tested the performance without query logs.  I disabled the query log feature , but  it seemed that QPS  couldn't  go higher (about 75KQPS)  when CPU still had lots of room  --  Named CPU Usage was about 250%,  but in theory  it can reach to 400% (4 vCPU).


In your opinion,  which default parameters I need pay attention to , which will impact system capacity performance when query log is off?


Our Bind version is 9.10.5.  Do you know what version you think will have better multi-threading ?


Thank you very much!

Michael

-----Original Message-----
From: PENG, JUNAN 
Sent: Wednesday, February 21, 2018 11:32 AM
To: 'Tony Finch' <dot at dotat.at>
Subject: RE: Help 

Another thing:

When query log is on.

Traffic is 35k QPS,   the Named CPU usage is 260%
Traffic is 40K QPS,  the Named CPU usage is 260% Traffic is 50K QPS,  the Named CPU usage is 260% Traffic is 70 -100 KQPS, the named CPU usage is still 260% 

When the named CPU usage has a limitation 260% here ? (in theory, it should be 400%)

BR
Michael

-----Original Message-----
From: Tony Finch [mailto:dot at dotat.at]
Sent: Wednesday, February 21, 2018 11:09 AM
To: PENG, JUNAN <jp2111 at att.com>
Cc: bind-users at lists.isc.org
Subject: Re: Help 

PENG, JUNAN <jp2111 at att.com> wrote:
>
> Why Query log off/on  feature is impacting named CPU Usage ?

It has to serialize query processing in order to write to the log, and that serialization barrier limits the parallelism that it can achieve (due to Amdahl's law).

Tony.
-- 
f.anthony.n.finch  <dot at dotat.at>  https://urldefense.proofpoint.com/v2/url?u=http-3A__dotat.at_&d=DwIBAg&c=LFYZ-o9_HUMeMTSQicvjIg&r=xVh2hygmrxdOcVtuPuDNjQ&m=6ZyClPvp9-p1nUQmEemmpxj_3-f1_sW8Fpnap_DuXjE&s=5383B5P89WllOA-wt8I4eVUaH2IU238-fa0ljJHGu38&e=   -  I xn--zr8h punycode
South Utsire: Northerly or northeasterly 4 or 5, becoming variable 3 or 4.
Slight or moderate. Fair. Good.