Billions Connected

Lifting the Fog Around the Stupid Network – iPlane

“Deliver the Bits, Stupid.”

- David Isenberg – Rise of the Stupid Network

Stupidity is a virtue – for networks.

But not all networks should be stupid – only the underlying IP network our applications are built on.

Case in point: the overlay networks implemented by today’s new breed of distributed applications are getting smarter and smarter – and this is a very good thing. But overlay networks have been severely limited by the fogginess of the stupid network.

The fog of the stupid network is unknown loss, unknown latency, unknown bandwidth. Packets get dropped into the ether and there only are a few strategies an application developer can turn to in dealing with issues of latency, loss or congestion. The common techniques revolve around passive mechanisms for dealing with network conditions as they occur. Jitter buffering and streaming media throttling fit this description. Good techniques – but not good enough.

The solution to this problem is not to get rid of the stupid network, but to map it. Lift the fog.

That’s what the Systems and Networking team at UW has done with iPlane. Instead of making guesses with path-local information (think ping, traceroute) about properties of the network that are essentially non-local, applications can make direct use of a global view of these properties.

iPlane consists of two pieces:

• A set of mechanisms for collecting network connectivity and path information and supplementing the resulting map with hints that implies routing (location, topology, and other details)
• A simple query interface to the resulting map database

The iPlane team was kind enough to provide me with access to an early version of the query interface – it’s rough around the edges but has huge potential for improving the performance of today’s most interesting new applications (Joost comes to mind).

iPlane is a leap ahead of the baby steps made so far towards active techniques for dealing with poor network conditions. Skype attempts to route voice traffic through relay peers – but as any Skype user knows the resulting call quality can still run the whole range from very good to very bad. Many P2P networks preferentially remain connected to faster hosts while disconnecting from under-performers – but this behavior, while able to improve download speeds, is based on simple heuristics and “converges” towards more optimal sets of download hosts fairly slowly. iPlane will enable application overlay networks to go far beyond these techniques.

Check out the excellent talk on iPlane by Tom Anderson on Google TechTalks.

Jeff

Tags: iplane, ipnetworking, iprouting, isenberg, stupidnetwork, tomanderson, universityofwashington, uw

Freebase – The Best Thing to Happen to the Creative Commons since Lawrence Lessig?

For those who don’t know what Freebase is yet, read Tim O’Reilly’s post on the topic first, then come back here.

Now, for the rest of you compulsive blog readers – consider Freebase vs. Google.

Google organizes information by creating indexes of that information – using the fair use exception to create those indexes. Freebase is different.

They’re pulling in all the CC-licensed material they can find on the web – hence the emphasis on the bootstrapping of their database with Wikipedia content.

This makes sense. Freebase can only do its semantic magic by making entire documents open to the web for the addition of semantic annotation. In other words – building indexes is not good enough anymore. If Freebase is to fulfill its purpose to bootstrap the semantic web it must exceed the bounds of fair use. Hence the necessity of fueling Freebase with CC-licensed content.

So does Freebase become a CC freeloader? Not at all.

In fact Freebase seems likely to give CC-licensed content a strategic advantage over non-CC-licensed content. Given a web enabled with meaning vs. the web of today, which would you rather surf? Traditionally copyrighted material that is harder to pull into the semantic sphere due to legal barriers could end up looking static and tired by comparison.

So it seems that Freebase may be able to catalyze a virtuous cycle of Creative Commons licensing.

This raises the possibility of a new intersection between copyright law and the web:

The success of Google and the search economy as a whole is driven off of the fair use exception of copyright law. Could it be that the CC licensing model turns out to be an a key enabler of Web 3.0?

It’s an intriguing possibility.

Jeff LaPorte
EQO Founder and Chief Architect

Tags: creativecommons, freebase, lessig, metaweb, semantic, semweb, wikipedia