Monday, December 26, 2005
Cache as Cash Can
I'm in Adelphia territory for the holidays. This morning it took 14 seconds for isen.blog to load. Funny, it only took the New York Times about three seconds. To better understand what gives here, I timed the following web pages from <CR> to the first screen of text:
hyperorg.com -- 4 sec
mediadiet.com -- 11 sec
saschameinrath.com -- 8 sec
freedom-to-tinker.com -- 7 sec
webblogg-ed.com -- 9 sec
wifinetnews.com -- 10 sec
Then, a few minutes later, I did it again:
hyperorg.com -- 1 sec
mediadiet.com -- 1 sec
saschameinrath.com -- 4 sec
freedom-to-tinker.com -- 1 sec
webblogg-ed.com -- 3 sec
wifinetnews.com -- 1 sec
For comparison
New York Times -- 3 sec
Boston Globe -- 3 sec
Newsweek -- 4 sec
Christian Science Monitor -- 5 sec
Mercury News -- 2 sec
So recency helps. So does popularity (i.e., somebody else's recency). It is a stupid cache, and as readers of isen.blog know, stupidity is good. But what if caching were based on some other more "intelligent" criterion, like, for example, suppose latency were the inverse of how much Ed Whitacre got paid? Then how would we feel if, say, isen.blog always took 14 sec to load?
Too many of us are stuck on a network neutrality argument that centers around the Googles and Yahoos and Amazons. But when the network neutrality tub is emptied, the big fish will survive. It'll be us little babies that get sucked down the drain.
hyperorg.com -- 4 sec
mediadiet.com -- 11 sec
saschameinrath.com -- 8 sec
freedom-to-tinker.com -- 7 sec
webblogg-ed.com -- 9 sec
wifinetnews.com -- 10 sec
Then, a few minutes later, I did it again:
hyperorg.com -- 1 sec
mediadiet.com -- 1 sec
saschameinrath.com -- 4 sec
freedom-to-tinker.com -- 1 sec
webblogg-ed.com -- 3 sec
wifinetnews.com -- 1 sec
For comparison
New York Times -- 3 sec
Boston Globe -- 3 sec
Newsweek -- 4 sec
Christian Science Monitor -- 5 sec
Mercury News -- 2 sec
So recency helps. So does popularity (i.e., somebody else's recency). It is a stupid cache, and as readers of isen.blog know, stupidity is good. But what if caching were based on some other more "intelligent" criterion, like, for example, suppose latency were the inverse of how much Ed Whitacre got paid? Then how would we feel if, say, isen.blog always took 14 sec to load?
Too many of us are stuck on a network neutrality argument that centers around the Googles and Yahoos and Amazons. But when the network neutrality tub is emptied, the big fish will survive. It'll be us little babies that get sucked down the drain.
Technorati Tags: Cableco, F2C, IntelligentNetwork, Internet, NetworkNeutrality, OpenAccess
Comments:
david, you undoubtedly thought of this already, but did you clear your browser's cache between the first and second times you loaded the pages?
# posted by 
: 12/26/2005 5:36 PM
: 12/26/2005 5:36 PM
I see your point that load times matter, and that user-experienced-latency is not just about tcp-layer issues.
However, it's hard to understand from the numbers you give WHICH type of caching you're measuring, and therefore what level of conspiracy you're worried will fail to cache unpopular or unpaid sites.
There's a lot of caching that goes on in the net, and most of it is private, on the part of the site or the end user. The telco (when it is not your ISP) probably does little caching.
Unless I'm mistaken. If so, please enlighten with a little more info on which caching is happening (or not) for big sites and not little ones.
However, it's hard to understand from the numbers you give WHICH type of caching you're measuring, and therefore what level of conspiracy you're worried will fail to cache unpopular or unpaid sites.
There's a lot of caching that goes on in the net, and most of it is private, on the part of the site or the end user. The telco (when it is not your ISP) probably does little caching.
Unless I'm mistaken. If so, please enlighten with a little more info on which caching is happening (or not) for big sites and not little ones.
# posted by : 12/27/2005 5:18 PM
: 12/27/2005 5:18 PM
The big fish for surviving still needs food.The end of the predators usually comes for lack of preys to hunt.
Nature is a balance where every living creature has its value, as eater as well as food.
If the little babies are sucked down the drain, the big fish will have to skip lunch and may be dinner too...
Patrizia
Nature is a balance where every living creature has its value, as eater as well as food.
If the little babies are sucked down the drain, the big fish will have to skip lunch and may be dinner too...
Patrizia
I held down the shift key as I clicked on the links. My understanding is that this bypasses browser cache, so that load times would measure network latencies. For fun, though, I explicitly cleared my browser's cache and tried again -- load times were comparable to the second set of numbers. If filling browser cache were a big contributor to delay, with an empty cache I would have seen data like the first data set. But I didn't so I conclude that the improvement in load times come from some cache in the network.
Re: Anonymous #2's comment, I am assuming, tentatively, that the facilities-based ISP that provided my cable connection, Adelphia, was doing the caching. I suppose it could have been anywhere in the network, but I agree that nobody else would have strong economic motivation to do it.
Post a Comment
Re: Anonymous #2's comment, I am assuming, tentatively, that the facilities-based ISP that provided my cable connection, Adelphia, was doing the caching. I suppose it could have been anywhere in the network, but I agree that nobody else would have strong economic motivation to do it.
