W3C home > Mailing lists > Public > www-tag@w3.org > June 2012

What would count as good measurements wrt the utility of SPDY?

From: Henry S. Thompson <ht@inf.ed.ac.uk>
Date: Wed, 13 Jun 2012 20:36:37 +0100
To: www-tag@w3.org
Message-ID: <f5bvcivqa0q.fsf@calexico.inf.ed.ac.uk>
Some discussion at the TAG f2f involved glancing reference to the
question of whether SPDY would improve browser (user) experience.

I did a _tiny_ experiment to just see what kind of numbers I would
find.  Here are the results.  I offer them more as a stimulus to
thinking about just how complicated it is to get a handle on the above
question, how many conditions have to be controlled for, etc., than
anything else.

I also am aware that there are serious efforts underway to do this
'properly' -- this is rank amateurism.

ht
------------------
Fetching www.weather.com on 2012-06-13 produces 119 requests according
to fiddler, and then starts polling for some updating flash frame(s).

Response codes are as follows:
    112 200
      1 204
      6 302
--------
Sites are as follows:
     34 i.imwx.com
     14 s.imwx.com
     12 www.weather.com
      9 ad.doubleclick.net                  ---    69 == 58%
      7 b.imwx.com
      7 d.imwx.com                          ---    83 == 70%
      6 s0.2mdn.net
      4 cp30559.edgefcs.net
      4 pix04.revsci.net
      2 admin.brightcove.com
      2 b.scorecardresearch.com
      2 odc.weather.com
      2 secure-us.imrworldwide.com
      2 speed.pointroll.com		    ---    107 == 90%
      1 x 12
So, exponential/Zipf, not normal
90% of the requests are to 54% (14 out of 26) of the hosts
At least *.imwx.com and edgefcs (manages flash polling) are akamai
-------
Traffice to the top 14 hosts measured in bytes

			n	mean	total	s.d.
i.imwx.com		34	6271.62	213235	8354.51	*
s.imwx.com		14	8070.86	112992	6678.07	*
www.weather.com		12	8742.17	104906	13313.3	*
ad.doubleclick.net	9	498.556	4487	526.193	
b.imwx.com		7	43	301	0
d.imwx.com		7	16518	115626	17250	*
s0.2mdn.net		6	4869.33	29216	5155.37
cp30559.edgefcs.net	4	72.5	290	81.0987
pix04.revsci.net	4	257.5	1030	193.736
admin.brightcove.com	2	12929	25858	5051.57
b.scorecardresearch.com	2	570	1140	806.102
odc.weather.com		2	21.5	43	30.4056
secure-us.imrworldwide.	2	93.5	187	54.4472
speed.pointroll.com	2	63357.5	126715	61188.1
12 more                 1	4004.58	48055	6225.55


Total                 119	6588.92	784081	12875

*ed lines say that 4 sites account for 56% of the requests and 70% of
the bytes

Number of requests and bytecount are roughly correlated (spearman rank
correlation .51, p<.01), with speed.pointroll.com being the obvious outlier)

-- 
       Henry S. Thompson, School of Informatics, University of Edinburgh
      10 Crichton Street, Edinburgh EH8 9AB, SCOTLAND -- (44) 131 650-4440
                Fax: (44) 131 650-4587, e-mail: ht@inf.ed.ac.uk
                       URL: http://www.ltg.ed.ac.uk/~ht/
 [mail from me _always_ has a .sig like this -- mail without it is forged spam]
Received on Wednesday, 13 June 2012 19:37:03 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 22:56:45 UTC