1DNSTCPBENCH(1)           PowerDNS Authoritative Server          DNSTCPBENCH(1)
2
3
4

NAME

6       dnstcpbench - tool to perform TCP benchmarking of nameservers
7

SYNOPSIS

9       dnstcpbench [OPTION]... REMOTE-ADDRESS [REMOTE-PORT]
10

DESCRIPTION

12       dnstcpbench  reads  DNS  queries  (by  default from standard input) and
13       sends them out in parallel to a remote nameserver. By default TCP/IP is
14       used,  but  optionally, UDP is tried first, which allows for the bench‐
15       marking of TCP/IP fallback.
16
17       The program reports both mean and median numbers for queries per second
18       and  UDP  and  TCP  latency. Each query only counts once, even if it is
19       tried over UDP first. This effectively  means  that  passing  '-u'  can
20       lower query rates if many queries get shunted to TCP.
21
22       The  input  format  is one query per line: qname single-space qtype. An
23       example:
24
25          www.powerdns.com ANY
26
27       When  benchmarking  extended  runs,  it  may  be  necessary  to  enable
28       TIME_WAIT  recycling,  as  TCP/IP port tuples may otherwise run out. On
29       Linux this is performed by running:
30
31          echo 1 > /proc/sys/net/ipv4/tcp_tw_recycle
32
33       The equivalent for IPv6 is not known.
34

OPTIONS

36       -f, <FILENAME>, --file <FILENAME>       FILENAME  from  which  to  read
37       queries.  Defaults  to  standard  input  if  unspecified.   -h,  --help
38       Provide     a     helpful     message.       --timeout-msec      <MSEC>
39       MSEC   milliseconds   to   wait   for   an   answer.   -u,  --udp-first
40       Attempt resolution via UDP first, only do TCP if  truncated  answer  is
41       received.  -v, --verbose                           Be wordy on what the
42       program is doing.  --workers <NUM>                         Use NUM par‐
43       allel worker threads.
44
45       REMOTE-ADDRESS: IPv4 or IPv6 to test against.
46
47       REMOTE-PORT: Port to test against, defaults to 53.
48

BUGS

50       Currently  the timeout code does not actually perform non-blocking con‐
51       nects or writes. So a slow connect or slow writes will still cause  low
52       performance and delays.
53
54       Median  queries  per second statistics are reported as 0 for sub-second
55       runs.
56

AUTHOR

58       PowerDNS.COM BV
59
61       2001-2019, PowerDNS.COM BV
62
63
64
65
664.2                              Nov 29, 2019                   DNSTCPBENCH(1)
Impressum