"I photographed every door or drawer knob, handle, or latch I touched
from the time I awoke on Thursday, June 3rd.
until I went to bed on Friday, June 4th"
That was a nice Fluxus piece. I went back to look at the site again, and got a page from Earthlink saying that the site had exceeded its free monthly traffic allotment, and would be available for viewing again AT THE BEGINNING OF NEXT MONTH! Talk about tough love--that's three weeks away! Is "traffic" really that expensive? I guess this is what you were explaining to me about e-commerce (or e-art), Jim: the price of success is failure.
Any reasonable amount of traffic is so cheap as to be basically free. The problem is that the network effect can potentially bring you unreasonable amounts of traffic. In our current set up we pay $25/month for 10 gigabytes of data transfer. After that we pay 10 cents a megabyte. If we assume pages are around 50k each (although the front page is much less, and pages with lots of pictures are more) then that means 20 page views = 1 meg of traffic, so 20,000 page views = 1 gigabyte of traffic, and 200,000 page views would equal our 10 gig limit. We would not be shut down at this point though, we would just start being charged. In fact there would be no real way to turn it off. I guess the problem is that if you get 200,000 page views a month then there's really no reason why you couldn't (or wouldn't) get a million, and something in that range would be expensive. (800,000 extra hits above the 200,000 limit x 50k = 40,000 megs of extra traffic which would be at 10 cents a megabyte x 40,000 = $4000 extra charge.) So yes, it can get expensive. But that's a ridiculous amount of traffic. We transfer in the neighborhood of 500 megs a month (1/2 a gigabyte out of our allowed 10.) Basically all it takes is for slashdot to have a story about you on their front page and your servers will go down with tens of thousands of simulataneous hits.
My philosophy is that there is no point in just blindly trying to accumulate hits. The economics have already proven there is no way to turn all those hits into anything but a revenue drain (because of the bandwidth.) And the problem only gets worse as the media gets more rich. That doorknob page was several hundred kilobytes (instead of 50) and something like a grainy little postage stamp sized movie video could be several megabytes. That stuff can really add up. I love the idea that the web is free for surfers, and I love the openness of these forums. But I don't consider it a goal just to get a lot of people to look. I guess the goal is more to get the right kind of people to look.
Thanks for walking me through the math--I've not seen it so clearly explained before. Slashdot, hell--think about what must have happened when Thomas Friedman of the New York Times did an enthusiastic column a couple of years back on that cottage e-business in the Midwest (or wherever it was). He basically signed their death warrant! (They went belly up a few months after the column, but Friedman never admitted his own role--in their initial success or later failure.) Actually, Friedman's been strangely silent these days after all his foolish e-commerce predictions. If he'd had you to explain the numbers, he not have crowed so loudly about the "internet economy."
But that Fluxus guy shouldn't have been shut down--most folks don't give a crap about avant garde art! Don't you think Earthlink is setting the bar a bit low?
You get what you pay for. A lot of hosts advertise plans like $20/month for UNLIMITED bandwidth (where bandwidth refers to total bytes transfered.) This is never ever true. Logically it can't be true because the host has only a finite amount of bandwidth (even when this is a very large number, as in the case of Earthlink.) Anyway, what these flat rate people do is just cut you off if you get too popular. With other (I think more reputable) hosts you pay a flat fee for anything up to a specific point, and then you pay by the byte (be it mega or giga in size.) Your potential losses are more in this system, but at least you are sort of in control and you know what's happening, you can always take the site off line temporarily and serve up a 2k page offering some explanation (and maybe a link to paypal or something.) But for some projects it makes sense to go with a provider who will just cut you off. That way there is no big financial downside if you get overly popular.
As for the Times guy, I wonder if it's true that the a NYTimes link would generate more hits than a slashdot link. I'm not so sure about that. They don't call it the "slashdot effect" for nothing.
One last thing which might be of interest is that as you consider larger and larger internet players the story sort of changes. The internet is really a collection of slightly smaller networks all joined together. These "smaller" (ha!) networks are things like MCI/Worldcom and Sprint and UUnet and the US Military, etc. (probably the companies are out of date, but I can't keep up with all the mergers and whatnot.) Anyway, for the big players, they've always had traffic sharing agreements. I'm big network company A and I want all my customers to have access to content on network company B's network, so we set up a traffic sharing arrangement. I'll let all of B's customers send data over my network and in return all of my customers can send data over company B's network. This is basically the reason why the internet seems free to the casual (low bandwidth) surfer. So the interesting thing is that at the top of the pyramid, having very popular content that a very wide audience is anxious to see, is valuable in that it gets you access to these traffic sharing agreements. I'm not suggesting that any one specific site could be big enough (well, something like Aol is big enough, but they're not really a site) to get this sort of deal, but it's interesting none the less. Being popular costs because someone has to pay for the bandwidth and it is not going to be the surfer. But being so popular as to be crucial in some sense means that all the other networks will pay for your bandwidth (not in terms of money, but in terms of giving you their bandwidth.) I can't fully understand this level of things, and of course these deals are only open to the mega huge corporate powers, but it might be that it's an anthropologically interesting arrangement (whatever that means.)
Actually it wasn't a Times link but a 600 word column on the Times Op Ed page, which goes to millions of surfers and non-surfers the world over. At the time old-media plugs (free or otherwise) were directing huge volumes of traffic to specific sites. Is the Times vs Slashdot the Hulk vs the Thing or the Hulk vs Spider-Man? I guess we could look it up.
Thanks for the info about traffic-sharing. More stuff I didn't know. It is anthropologically interesting because it seems to say more about status and old-boy networks than economics.
Here's a good metafilter thread about bandwidth and the way these costs can suddenly balloon when a site gets popular. Baseballprimer.com is the site in question. Right now their front page reads (in part): "We apologize for the inconvenience, but due to a large amount traffic,
Baseball-Reference.com has been temporarily disabled by our hosting provider
Communitech.net. While lots of traffic is generally not thought of as a bad thing,
"unlimited bandwidth" does not always mean unlimited bandwidth. In the first week
of April, Baseball-Reference served 1.07 million files (134k per day) and a total of
650k in individual pages (85k of these are scripts that require a trip to the
database), while servicing 42,000 distinct hosts. The 11.3 GB of data transferred is
also a very high total. We hope to have the site up and running very soon (maybe
tomorrow), but these issues are unlikely to go away and may only become more
acute as the site continues to gain in popularity."
The metafilter thread has more links to other interesting things on this topic. Before anyone starts thinking that our fun here could be in jeapordy, just keep in mind that 11.3 gigs is A LOT of data to be serving per month. So far this month (april 1 through april 12) we have served approximately 170 megs of information. We can double in size four times before any of these concerns come into play. And with my crappy database programming you can be sure that the site will break down long before that.
news
|
- jim 4-07-2001 4:37 pm
That was a nice Fluxus piece. I went back to look at the site again, and got a page from Earthlink saying that the site had exceeded its free monthly traffic allotment, and would be available for viewing again AT THE BEGINNING OF NEXT MONTH! Talk about tough love--that's three weeks away! Is "traffic" really that expensive? I guess this is what you were explaining to me about e-commerce (or e-art), Jim: the price of success is failure.
- Tom Moody 4-09-2001 5:40 am [add a comment]
Any reasonable amount of traffic is so cheap as to be basically free. The problem is that the network effect can potentially bring you unreasonable amounts of traffic. In our current set up we pay $25/month for 10 gigabytes of data transfer. After that we pay 10 cents a megabyte. If we assume pages are around 50k each (although the front page is much less, and pages with lots of pictures are more) then that means 20 page views = 1 meg of traffic, so 20,000 page views = 1 gigabyte of traffic, and 200,000 page views would equal our 10 gig limit. We would not be shut down at this point though, we would just start being charged. In fact there would be no real way to turn it off. I guess the problem is that if you get 200,000 page views a month then there's really no reason why you couldn't (or wouldn't) get a million, and something in that range would be expensive. (800,000 extra hits above the 200,000 limit x 50k = 40,000 megs of extra traffic which would be at 10 cents a megabyte x 40,000 = $4000 extra charge.) So yes, it can get expensive. But that's a ridiculous amount of traffic. We transfer in the neighborhood of 500 megs a month (1/2 a gigabyte out of our allowed 10.) Basically all it takes is for slashdot to have a story about you on their front page and your servers will go down with tens of thousands of simulataneous hits.
My philosophy is that there is no point in just blindly trying to accumulate hits. The economics have already proven there is no way to turn all those hits into anything but a revenue drain (because of the bandwidth.) And the problem only gets worse as the media gets more rich. That doorknob page was several hundred kilobytes (instead of 50) and something like a grainy little postage stamp sized movie video could be several megabytes. That stuff can really add up. I love the idea that the web is free for surfers, and I love the openness of these forums. But I don't consider it a goal just to get a lot of people to look. I guess the goal is more to get the right kind of people to look.
- jim 4-09-2001 5:47 pm [add a comment]
Thanks for walking me through the math--I've not seen it so clearly explained before. Slashdot, hell--think about what must have happened when Thomas Friedman of the New York Times did an enthusiastic column a couple of years back on that cottage e-business in the Midwest (or wherever it was). He basically signed their death warrant! (They went belly up a few months after the column, but Friedman never admitted his own role--in their initial success or later failure.) Actually, Friedman's been strangely silent these days after all his foolish e-commerce predictions. If he'd had you to explain the numbers, he not have crowed so loudly about the "internet economy."
But that Fluxus guy shouldn't have been shut down--most folks don't give a crap about avant garde art! Don't you think Earthlink is setting the bar a bit low?
- Tom Moody 4-09-2001 8:30 pm [add a comment]
You get what you pay for. A lot of hosts advertise plans like $20/month for UNLIMITED bandwidth (where bandwidth refers to total bytes transfered.) This is never ever true. Logically it can't be true because the host has only a finite amount of bandwidth (even when this is a very large number, as in the case of Earthlink.) Anyway, what these flat rate people do is just cut you off if you get too popular. With other (I think more reputable) hosts you pay a flat fee for anything up to a specific point, and then you pay by the byte (be it mega or giga in size.) Your potential losses are more in this system, but at least you are sort of in control and you know what's happening, you can always take the site off line temporarily and serve up a 2k page offering some explanation (and maybe a link to paypal or something.) But for some projects it makes sense to go with a provider who will just cut you off. That way there is no big financial downside if you get overly popular.
As for the Times guy, I wonder if it's true that the a NYTimes link would generate more hits than a slashdot link. I'm not so sure about that. They don't call it the "slashdot effect" for nothing.
One last thing which might be of interest is that as you consider larger and larger internet players the story sort of changes. The internet is really a collection of slightly smaller networks all joined together. These "smaller" (ha!) networks are things like MCI/Worldcom and Sprint and UUnet and the US Military, etc. (probably the companies are out of date, but I can't keep up with all the mergers and whatnot.) Anyway, for the big players, they've always had traffic sharing agreements. I'm big network company A and I want all my customers to have access to content on network company B's network, so we set up a traffic sharing arrangement. I'll let all of B's customers send data over my network and in return all of my customers can send data over company B's network. This is basically the reason why the internet seems free to the casual (low bandwidth) surfer. So the interesting thing is that at the top of the pyramid, having very popular content that a very wide audience is anxious to see, is valuable in that it gets you access to these traffic sharing agreements. I'm not suggesting that any one specific site could be big enough (well, something like Aol is big enough, but they're not really a site) to get this sort of deal, but it's interesting none the less. Being popular costs because someone has to pay for the bandwidth and it is not going to be the surfer. But being so popular as to be crucial in some sense means that all the other networks will pay for your bandwidth (not in terms of money, but in terms of giving you their bandwidth.) I can't fully understand this level of things, and of course these deals are only open to the mega huge corporate powers, but it might be that it's an anthropologically interesting arrangement (whatever that means.)
- jim 4-09-2001 8:57 pm [add a comment]
Actually it wasn't a Times link but a 600 word column on the Times Op Ed page, which goes to millions of surfers and non-surfers the world over. At the time old-media plugs (free or otherwise) were directing huge volumes of traffic to specific sites. Is the Times vs Slashdot the Hulk vs the Thing or the Hulk vs Spider-Man? I guess we could look it up.
Thanks for the info about traffic-sharing. More stuff I didn't know. It is anthropologically interesting because it seems to say more about status and old-boy networks than economics.
- Tom Moody 4-09-2001 9:28 pm [add a comment]
Maybe Hulk vs. Reed Richards
- jim 4-09-2001 9:40 pm [add a comment]
Here's a good metafilter thread about bandwidth and the way these costs can suddenly balloon when a site gets popular. Baseballprimer.com is the site in question. Right now their front page reads (in part):
The metafilter thread has more links to other interesting things on this topic. Before anyone starts thinking that our fun here could be in jeapordy, just keep in mind that 11.3 gigs is A LOT of data to be serving per month. So far this month (april 1 through april 12) we have served approximately 170 megs of information. We can double in size four times before any of these concerns come into play. And with my crappy database programming you can be sure that the site will break down long before that.
- jim 4-13-2001 8:48 pm [add a comment]
So if we post 4 X as many pictures, we can tip this puppy over !?
I'm sorry, it sounded like a challenge.
- bill 4-13-2001 9:01 pm [add a comment]
news
- news- 8-09-2004 11:22 am [add a comment]