Estimating DNS bandwidth: Average size of a DNS request?
I'm trying to estimate the amount of traffic 20,000,000 DNS requests a month will generate in terms of bandwidth. I'll include my calculations below but they all hinge on my estimation of the amount of data used to make a simple A request to a DNS server.
From my tests I think that 50 bytes is about right but wondered if anyone knew differently.
My calculations based around 50 bytes per query:
520 bytes for a DNS request 200000000 DNS monthly requests 104000000000 monthly bytes 832000000000 monthly bits 27733333333.33 daily bits 320987.65 bits/s 313.46 kb/s
Thanks for getting this far!
I think your data needs some new approximations, since a usual DNS server reply is smaller than 520 bytes (in fact, most of the routers (or networking equipment) can give you headaches when the UDP packet size passes 512kb mark - but we're not talking here about only UDP).
Here it goes - will use two very known linux tools to approximate the size of a typical DNS request.
$ dig linux.org +stats
; <<>> DiG 9.6.1-P1 <<>> linux.org +stats
;; global options: +cmd
;; Got answer:
;; ->>HEADER<<- opcode: QUERY, status: NOERROR, id: 7061
;; flags: qr rd ra; QUERY: 1, ANSWER: 1, AUTHORITY: 2, ADDITIONAL: 0
;; QUESTION SECTION:
;linux.org. IN A
;; ANSWER SECTION:
linux.org. 43200 IN A 198.182.196.48
;; AUTHORITY SECTION:
linux.org. 43180 IN NS ns0.aitcom.net.
linux.org. 43180 IN NS ns.invlogic.com.
;; Query time: 239 msec
;; SERVER: 127.0.0.1#53(127.0.0.1)
;; WHEN: Thu Oct 29 11:52:44 2009
;; MSG SIZE rcvd: 100
As you see, I made a DNS query to the local DNS server, loopback interface (for simplicity and clarity). You should find interesting the last row "MSG SIZE"...
Confirm it with tcpdump (running on loopback interface) :
IP localhost.36855 > localhost.domain: 7061+ A? linux.org. (27)
IP localhost.domain > localhost.36855: 7061 1/2/0 A 198.182.196.48 (100)
What you see in the end of each row it's the actual size (the thing you're looking for).
I advise you to run several test queries and average your DNS request size in your calculations. Keep an eye out for domains that aren't directely served from your DNS servers (that should be an interesting bit).
Kaplah.