[dns-operations] What are the data? [was: After Google Mail, Google Docs, Google Wave... Google DNS]

Patrick W. Gilmore patrick at ianai.net
Mon Dec 7 18:57:23 UTC 2009


On Dec 7, 2009, at 12:26 PM, Paul Vixie wrote:

>> From: "Patrick W. Gilmore" <patrick at ianai.net>
>> Date: Mon, 7 Dec 2009 07:29:17 -0500
>> 
>> What are the _data_?  While name calling on mailing lists is easy and
>> fun, it is not data.
> 
> see <http://en.wikipedia.org/wiki/Burden_of_proof> and consider that the
> status quo is a phat web farm and load balancer, whereas the benefits of
> wide area distribution and the use of dns tricks to pre-route consumers
> to one mirror vs another is the thing for which data needs be supplied.
> (and i'd want that data to include a comparison to the ibm websphere
> approach where the comsumer-specific routing is done with web redirects
> rather than with dns.)

I made no claims about CDN performance vs. websphere.  In fact I made no claims at all.  I intentionally ignored the specific case at hand and never even mentioned Google or CDNs.

My only point was that instead of name calling, we should look at data when judging the efficacy of an algorithm.  While it is true the Internet is full of children, we don't need to be childish about it.


>> If the real-world, objective, empirical data show a performance or other
>> gain, then the algorithm is not "silly", and no number of posts from
>> anyone no matter how famous will matter.
> 
> akamaized web sites do work better, because of various disciplines applied,
> like relative links, w3c compliance, and similar.  (this is my actual
> experience.)  but that's not the only benefit claimed by CDN providers.

I repeat: Ask the consumers of the algorithm.

Taking CDN services as a hypothetical algorithm to be judged (despite my attempt to keep this more generalized), I'm certain many companies have put a website on a CDN (not just "akamaized" - there are other CDNs ya know) without changing the underlying code and measured the impact.  If you are not one of these companies, your "actual experience" is not directly on point.  You cannot and do not control the variables, the owner of the content does.

Companies are spending over a billion USD / year on CDN services.  Perhaps they are doing it because of shiny marketing, or because the sales people buy them steak dinners, or whatever.  But perhaps - just perhaps - not every corporate decision maker with a fat checkbook is an idiot or a sheep.  Perhaps they are looking at metrics, seeing what their business return is on that investment.  Especially as it is hypothetically possible the hypothetical service under consideration is more expensive than the status quo.

Data exists.  Maybe even someone here is using a CDN and can provide data.  Then we can make an informed decision.  Until then, posturing and name calling is ... silly. :)

Hypothetically, of course.

-- 
TTFN,
patrick




More information about the dns-operations mailing list