Mojo Nation Respondsby Jim McCoy
Too cheap to meter?
Is the digital pasture we share endless or not? If it is -- when bandwidth, disk space, and processing cycles are really "too cheap to meter" or, in other words, "not worth worrying about" (how many people said electricity was too cheap to meter?) -- then resource contention is not an issue. However, if there is a boundary around this pasture, or if there are local bottlenecks, then the users must have a mechanism for allocating the resources.
Some things can be infinitely shared, like Mom's collection of digitized baby pictures. Some things cannot, like Mom's 500-MB hard drive. Mojo Nation's core technology handles each situation as efficiently as it can with its micropayment system, which also serves as an automatic distributed load balancer. In the case of Mom's baby pictures, demand probably won't be great, and the cost of providing that data is close to zero. Resource pricing and a payment system is only used when necessary -- if Mom's hard drive contains oft-requested content, then Mojo Nation's distributed load balancing moves some clients to a less-occupied server, while other users can use accumulated credit to move to the head of line.
In Clay Shirky's recent essays In Praise of Freeloading and Peers not Pareto, he writes that there is no "tragedy of the digital commons" -- where system performance degrades in relation to the weight of users who graze the system without returning anything to it -- because users don't care if their connection is tied up while they're not using it. For example, SETI@Home participants share their resources while they're not working.
In this case, the tragedy of the digital commons applies primarily to the downloading side of peer-to-peer service, the folks who wait in line to use Mom's 500-MB disk. If a user wants to download something but the only peers serving that content are overloaded, that user might want the option to pay the system to cut ahead in line.
In Mojo Nation, the micropayment system is just another tool we keep in our toolbox: The scorekeeping mechanism enables us to implement a peer-to-peer credit and reputation system, and handle distributed trust management and swarm distribution of data. It is also a flexible incentive device -- if the system keeps track of what each user contributes, then it is possible to reward the contributors. It is trivial to take a micropayment system and use it in Napster-like fashion -- if a Mojo Nation user sets his price for selling a block of data to zero, he has turned himself into a Napster-like peer.
Mojo Nation technology employs a market-based economy -- the shoppers and vendors make the rules, and they don't play with kids who lie or cheat -- to perform some accounting and management tasks that would otherwise fall upon centralized servers. The details can be completely invisible to the users -- Mojo Nation is just using a tool it has available, not pronouncing its treatise on sociology or politics.
Revisiting the digital commons
Napster, Shirky's favorite example of a digital cow pasture, is grazed by a certain kind of cow -- those who trade MP3 files swiped from a commercial compact disc. MP3 files are small and easy to transmit, even through a dial-up Internet connection (the kind employed by 85 percent of Internet users), and the record companies have done most of the indexing work. As long as this is true, Napster will enjoy what Dan Bricklin calls the cornucopia of the commons, where "use brings overflowing abundance."
During off-peak hours and other times of low network activity, all resource requests are treated equally, and Mojo Nation can operate in the same overflowing, abundant fashion as Napster. However, when resources are a source of contention and scarce, then a flexible payment/load balancing system shows its value.
Napster thrives as a narrow content pool, but open-ended content pools are a different playing field. Amidst a content collection without significant overlapping interest, there is lesser benefit for holding data for someone else -- that is, a user will be far more likely to share his disk space with the community if he is storing popular music rather than digitized baby pictures. In an open content pool, there is greater incentive for freeloaders, and the costs are borne by the honest users.
Free Riding on Gnutella, a research paper by Eytan Adar and Bernardo A. Huberman of Xerox PARC, revealed that 70 percent of the users in such systems provide no resources, leaving 1 percent of the peers to carry 50 percent of the load. The facts are clear.
If the shared resource is limitless, then there is never a problem -- we may all share and download until the, er, cows come home, and no one has to deal with the matter of resource contention. Even Napster is not so vast, though; Shirky says he leaves Napster running over a weekend, with the net result that the downloaders are better off while he is no worse off, but that is a shortsighted view. What about the unlucky downloaders stuck in line waiting for Shirky's mythical content? Mojo Nation is designed to cope with local resource contention problems like these, which arise naturally within distributed networks.
Again citing Shirky's example, a user on a broadband line can easily provide an MP3 file to a dial-up user, but the broadband user would find trying to download from the dial-up user intolerable. Mojo Nation's swarming distribution method enables multiple users -- even multiple dial-up users -- to contribute a small part of the work needed to deliver a large file to a requester, after which each worker bee collects credit according to his contribution to the group effort.
Additionally, the requesting user will not be delayed by a slow bee more than once. If the requester is in a hurry, the software agents working for him will shop for efficient service, passing by the slow vendors. That type of quality of service functionality is made possible once the resource allocation issues are handled -- in fact, it was Andrew Odlyzko's point in his Paris Metro Pricing paper that the simplest quality of service mechanism is one that enables users to pay a flat rate for a standard level of service and gives them the option to pay a little more to get on the fast track.
The underlying value of Mojo Nation's method of keeping score is flexibility. In situations where available resources are great, the internal incentives encourage multiple peers to work together at a task for optimum speed. In situations where resources are overloaded, clients are given incentives to switch to lesser-used servers.
Whether or not freeloading and cheating is a critical problem for P2P networks is an argument we will leave to academics and pundits such as Clay Shirky and the researchers at Xerox. When it comes to actually building a peer-driven distribution infrastructure, some choices need to be made.
In the design and implementation of the Mojo Nation technology, we decided to add those features that increased the security, flexibility, and scalability of our system. We have not been disappointed by any of our choices.
Jim McCoy is the founder and CEO of Mojo Nation.
Return to the P2P DevCenter.