Slashdot Mirror
Bees Can Optimize Internet Bottlenecks
prostoalex writes "Georgia Tech and University of Oxford scientists claim bees can help up develop a better Internet traffic algorithms. By observing bees, the researchers noticed that bees pass back information on route quality. 'On a basic level, the honeybee's dilemma is a tale of two flower patches. If one patch is yielding better nectar than the other, how can the hive use its workforce most efficiently to retrieve the best supply at the moment? The solution, which earned Austrian zoologist Karl von Frisch a Nobel Prize, is a communication system called the waggle dance.' Any practical applications of that? Well, apparently ad servers, serving banners across a variety of servers, can report back on the time it took to generate the page."
You know the tubes thing is a joke, right? You can't send live bees down it to figure out how to make it faster.
-----BEGIN PGP SIGNATURE-----
12345
-----END PGP SIGNATURE-----
Man I'm not even going to read the summary or TFA no time for that time to hire a ton of bees! Forget outsourcing to China and India and Eastern Europe, teh beehive is where it's at.
I like basketball!!1!
Well, I notice that one way of optimising _my_ experience is to switch off the ads, java, javascript and plug-ins. Blazingly fast, the tubes then become.
Every experiment which ends in a big bang is a good experiment.
I used to have a problem with bees around there the FIOS ONT cabling enters my house.
Freakin' hazard going anywhere near the thing.
And my internet is freakin' fast. They might be onto something.
Internet researchers get hives investigating honey of an optimization tool
Interested in a Flash-based MAME front end? Visit mame.danzbb.com
From November
-USR1
The problem of internet optimization stopped being a research subject years ago and is more of a business problem. Peering relationships and lowest cost routing mean that traffic will often travel a suboptimal route from a networking perspective because it is the best route from a providers financial perspective.
There are 4 boxes to use in the defense of liberty: soap, ballot, jury, ammo. Use in that order. Starting now.
Bees make the best web 2.0 buzzwords. They're the buzziest! I blame the hive mentality of the marketing drones for this.
today is spelling optional day.
"Beer Can Optimize Internet Bottlenecks"
I'm not sure what the alcohol has to do with network optimization, but I'll just say right now that I'm for it.
Is it just my observation, or are there way too many stupid people in the world?
The end of the article comments on how the bee algorithm also allows the data center to save power. It seems like if the data center isn't getting many requests, it can send some servers into power saving standby mode. Much like if there isn't much honey to collect, the bees stay in the hive and save energy.
Are other web data centers able to shut down some servers at night to save power, or is it just because this data center seems configured to allow the servers to each serve any number of websites?
What I got from reading the article was that they weren't optimizing the 'net at large but the services in one data center. By making the individual servers in the data center allocate themselves to the various hosted sites/services based on demand. Because of this, it's basically immune to external cheating, after all there is no point. If they were changing who's packets went through their network on the way to somewhere else, there would be a reason to cheat.
The article makes perfect sense, but the domain seems a little limited to me. You have to be able to quantize things. You have to be able to shift things around (make server A be able to pick role X, Y, or Z based on which is better at the moment). In some problem sets this would be easy. For example the /. setup that was described a while back where they have a few boxes doing this, a few doing that, and they all work off the same read only NSF share. It would be easy to move the boxes that run user pages between that and static pages. It could help there.
On the other hand, the boxes couldn't switch between being web and DB boxes very fast (you would have to load up all that data) so you couldn't let the boxes choose between those two roles (you'd lose most all your benefit from the expense of the switch).
The choices have to be relatively homogeneous.
Comment forecast: Bits of genius surrounded by a sea of mediocrity.
Why is it that the latest scientific research is always first applied to:
1) war
2) advertising on the net
If you've ever "herded" bees, you would know that even if you have a full suit, they'll attack you and be successful if you agitate them. So too with us nerds, if we figure out (and it's quite easy) that you're herding us to a specific patch for ad-revenue or whatyouwant, it will sting no matter who you are and what protection measures you have.
Custom electronics and digital signage for your business: www.evcircuits.com
Now we'll have routers do the wiggly bee dance
If a big Cisco or Juniper is running WDRP (Wiggly Dance Routing Protocol), does each line card do a different dance from the other cards in the chassis? What would such a router be called then, a hoedown? If the line cards started dancing "Thriller", will the router turn white and start chatting to young boys on IM?
In Soviet Russia, Jesus asks: "What Would You Do?"
Beers and bottlenecks? I think I will!
"Flyin' in just a sweet place,
Never been known to fail..."
The blurb is very misleading, this isn't about serving ads faster, or even about banner ads reporting their load times. It isn't about routing bottlenecks either, it is about distributed loads, and dynamic resource allocation on the server side.
The methodology used is this:
you have a server farm, this server farm is serving up many different sites. Internal to the server farm is an "ad board" for lack of a better term. When Site A's load spikes it's dedicated server can post an ad to the "ad board" which other servers in the farm can see. Then, servers which are dedicated to other sites, but are basically sitting idle can pick up the ad, say "oh I can help out this site over here" and somehow join the load balancing cluster that is server Site A's content. If necessary, the second (and however many other servers) can also place an ad on the board, getting more and more servers included serving up Site A.
As Site A's traffic decreases, less and less servers will be needed, so they will stop posting ads, and fewer and fewer servers will be serving Site A.
This is about dynamically allocating resources across a large data center/cluster not serving ads on the internet or even about optimal routing of traffic on the internet, instead of having a single server dedicated to serving 1 site, you have many servers which dynamically based on load decide which sites to serve.
African or European?
Ants also do a similar thing. The difference is that they release chemicals as they travel. The more ants that travel the same path, the stronger the trail, the more likely the path will continue to be traveled. This also has some consideration for traveling velocity. All other things being equal, a faster (typically shorter) path will have a stronger chemical signature because more ants will traverse it per unit time. Further, new, potentially better, paths will be discovered on rare occasion that an ant gets lost or for some reasons falls off the established path. In artificial simulations this trailblazing can be amplified by increasing the random chance that from any given network hop, the ant/packet will choose some new direction which may be totally random or may be based on other paths that have already been traversed.