You may not have heard of peer-to-peer processing, but you've surely heard of the controversial music site, Napster.com. The site made headlines last year through its use of peer-to-peer processing to enable Internet users to download and share music for free.
Though Napster's business practices may have been radical, peer-to-peer processing, or P-to-P, is anything but new. However, the Internet is giving new life to the idea of allowing personal computers to talk directly to one another (as "peers")-and venture capitalists are taking note. P-to-P offers Internet users a new way to communicate and collaborate for business.
A new breed of vendor, honing P-to-P for business use on the Internet, is emerging. Some have unnamed financial clients: capital markets players such as brokers, investment banks and fund managers. Others, including Berkley, CA-based Consilient, have clients running financial applications, such as e-billing. Consilient only released its first product at the start of the year.
Peer-to-peer processing is something bankers might consider as "the other P-to-P"-as opposed to person-to-person payments (which you'll see covered in our special payments report, starting on page 25.)
P-to-P computing encompasses P-to-P payment models and represents a much broader set of ideas than those at issue in the music companies' litigation against Napster. While noteworthy for its reach and popularity, Napster is driven by only one of four main applications of peer-to-peer technology.
The principle behind all types of P-to-P-the fact that computers can communicate directly, exchanging information or sharing power without a server to manage them-has been around for years. And the implications of the technology, particularly for banks, extend well beyond the bounds of just sharing files.
The emergence of Napster, along with other P-to-P-based systems such as Gnutella, ScourNet and SETI, may signal the beginning of a renaissance for peer-to-peer computing. Financial services companies, meanwhile, need look no farther than PayPal.com-the Palo Alto, CA-based online payment system-to see P-to-P at work in their own backyard.
But these consumer-focused applications, which have varied in popularity and financial success, are just the tip of the iceberg. A whole new breed of companies is honing in on the development of P-to-P computing for ecommerce, and they could change forever the way the many large companies, including financial firms, do business.
"Venture capitalists are foaming at the mouth over the possible application of P-to-P computing to the business world," says Simon Yates, an analyst with Forrester Research Inc., Cambridge, MA. In a research brief from the firm, Yates writes: "Intel is leading the high-profile P-to-P charge today, but by year-end a flurry of spirited startups will launch new products."
Despite Napster's high visibility, P-to-P computing is not synonymous with file sharing. At least four different types of P-to-P are emerging, says Yates, and among businesses, file sharing is probably not the most potentially lucrative capability.
While these technology models vary widely in objective and approach, they all embrace to some degree the notion of computers acting as peers to accomplish a task.
Among consumers, of course, the file sharing and advanced search applications performed by Napster and ScourNet are perhaps the most widely recognized form of P-to-P computing. But Yates points out that vendors such as Aimster and Centrata are "developing technology to perform natural language searches of millions of peer systems that will crush today's centralized searching technology from Inktomi and Google."
Perhaps even more relevant to the data-rich-and computationally heavy- financial services industry are a second class of P-to-P players: the "CPU- sharing" vendors. And just as the central processing unit, or CPU, essentially is the computer, these companies have created software and services aimed at tapping multiple computers' power to fuel processing- intensive efforts. It works like this: A software agent sitting on a PC notes when the computer is not being used, then lends its dormant CPU to a wider net of computers, all working on the same project. Working together, a large enough array of computers can muster as much computing power as some supercomputers.
SETI, or the Search For Extraterrestrial Intelligence at Home Project, is one interesting example of this CPU-sharing system at work. Sponsored by the University of California at Berkeley (setiathome.Berkeley.edu), the
program enlists volunteers who download an application file to their computers. The file enables project scientists to use the volunteers' computer power in processing radio trans-
missions from high-powered telescopes that scan the universe in search of intelligent life. Every few days, each computer in the program downloads data from the telescope. When
not in use by its owner, the computer processes the data for patterns and returns the results for analysis. With more than 2 million registered users, the pro ject has performed the largest cumulative computation to date, its managers say.
Again, bankers needn't look far to see essentially the same P-to-P capabilities at work in a financial application. New York City-based DataSynapse is already targeting the financial industry with its own CPU- sharing application. The company says its technology can help banks and brokerage firms become more efficient in handling computer-intensive chores such as risk analysis.
Unlike SETI, however, DataSynapse does not employ a "hub and spoke" model in which "I just send you data, you crunch the data and send it back," says the company's chief executive officer, Peter Lee. Instead, DataSynapse uses a purer form of P-to-P computing, he says, coupled with broadband resources to prevent data-processing logjams and speed computation even more.
Lee says the idea of "borrowing" CPU power from unused machines isn't really new to the financial services industry; but, until now, the machines needed to be dedicated to that purpose. He recalls that in walking through the capital markets unit of a large securities firm, "you would pass through and see a lot of stick-up notes saying 'this machine is idle'. If you'd moved it, or one machine crashed, the whole process would have to start over."
Lee says his company already has signed seven large securities firms as clients, as well as a number of major independent software vendors who plan to bundle DataSynapse's software into their middleware applications. With his company's software, Lee says, portfolio or hedge-fund managers' most unwieldy and time-consuming pricing models can be cut into small pieces, with each piece being distributed to any idle machines within the firm's operations. The software also manages potential breakdowns in the process: If an idle machine is needed, the piece of the task it is processing is moved automatically so the whole computation won't be lost. Lee claims the DataSynapse system can take a computing job that would normally consume between five and 70 hours and pare the job down to between 30 seconds and three minutes. The payoff for financial managers is faster turnaround on their assessments of the markets and thus faster action in response to those analyses.
"Think of a job where you have 1,000 tasks, each taking three minutes," says Lee. "Now, distribute that over 1,000 machines. That's how we can bring these large tasks back so quickly."
Although the ideas of peer-to-peer computing and even CPU sharing are not new, Lee says they have only now become commercially feasible due to three factors: first and foremost, a tremendous leap in desktop computing power in recent years, making even relatively inexpensive computers into powerful tools. For example, says Lee, "Nintendo came out with a game chip last year that was more powerful than the supercomputer Cray was selling for $8 million in 1986." In addition, steady gains in network bandwidth, along with the availability of object-oriented programming such as Sun Microsystems' Java, have provided the speed and modular building blocks necessary to make such applications viable.
Lee acknowledges one obstacle some financial services firms may encounter in such a P-to-P system: If computing tasks are written in "spaghetti code"-strings of flat programming, without the multi-tiered richness of object-oriented languages-and if the tasks are tightly interwoven into a database, it's very difficult to reconfigure the system to handle P-to-P architecture. Consequently, Lee says, DataSynapse and companies like it are focused on risk and portfolio management rather than settlement or transactions.
A third type of P-to-P computing that may attract bankers' attention is collaboration. Unlike the file-sharing or CPU-sharing models, P-to-P collaboration systems do not focus so much on borrowing something-be it existing data or unused computing power-as on connecting people and doing so in an interactive, real-time environment.
One of the early leaders in P-to-P collaboration is Groove Networks, a company founded by Ray Ozzie, the creator of the original Lotus Notes. Launched more than three years ago, Groove Networks introduced its peer computing platform this past October. Within six weeks, the Beverly, MA- based company had forged alliances with more than 50 systems integrators and third-party developers, including Perot Systems, which has several clients in the financial services industry. Groove's new "partners" are expected to use the company's platform to fashion their own peer-computing applications and solutions.
Groove works like this: A user downloads a 10-megabyte Groove Networks client, which they use to create a "shared space." Using participant lists- akin to the "Buddy Lists" used for instant messaging on America Online's service-other participants are invited to the shared space. They can accept or decline to join (unlike Napster's system, in which a user can't opt out of a file-sharing community). Once the invitation is accepted, all accepting clients can share files, create presentations and hold "live" meetings and discussions.
Andrew Mahon, director of strategic marketing for Groove, sees this as "more of a first cousin, once removed, from (Lotus) Notes" than a direct descendant. "The reason Ray founded Groove was that he was looking at a whole new style of computing," says Mahon.
The company's P-to-P technology can be especially valuable for large organizations, which increasingly choose to rely more heavily on outsourcing of one kind or another-consultants, freelancers, application service providers. Mahon says these companies need a way to reach beyond their centralized systems and enable collaboration among employees and outsiders, while not opening up their entire database.
Another company that fits into this segment is New York-based 3Path Inc., which promises to combine the richness of a Web site with real-time delivery of email to create a "third path" to customers. This implementation is closer to a traditional client-server model than Groove's offering or other P-to-P systems. While focused primarily-for now at least- on one-way information, 3Path CEO Brian Smiga says the architecture will support many-to-many collaboration. 3Path also draws on another once-lauded computing style, leveraging "push" vendor BackWeb's so-called "polite" push technology. It allows 3Path to download large files across uneven bandwidths without disrupting other applications or tying up the system.
Smiga says the technology is being tested by a number of unnamed beta clients, including a market research firm, several venture capital and recruiting firms, and at least one investment bank. The CEO believes a platform like 3Path could be extremely useful for delivering financial statements or presenting bills. 3Path will succeed where many push vendors have stumbled, he says, because its system is based on asking for users' permission, personalizing the information, and not tying up access and system space like platforms such as PointCast did.
Smiga also thinks his company has an edge on more radical platforms like Groove Networks' because 3Path still relies on the same basic rules and procedures as email or Web surfing and doesn't require the "paradigm shift that Groove is asking people to make."
A fourth area of P-to-P computing is business process coordination, which also focuses on enabling people to share information across systems. Unlike collaborative systems, these platforms and applications are targeted at easing the person-to-person interactions that come up during automated supply-chain processes, such as managing delivery schedules. Forrester's Yates said companies pursuing this technology include Consilient and Interbind.
Berkeley, CA-based Consilient creates "sitelets" that connect multiple applications into a core architecture so information can be shared by various parties, which can then accomplish tasks together. The sitelet is small, only 180K or so, and stores records of steps in the process so users can track an interaction throughout the supply chain. A major energy company is piloting the Consilient platform for their financial forecasting.
Here again, a company executive notes that the system is not based on a pure P-to-P computing model. Erik Freed, Consilient's chief technology officer, says the company's platform plugs into client-server architectures.
Most P-to-P business applications must be hybrids of sorts, at least initially, in order to work in tandem with legacy systems. What's more important is that the platforms embrace a P-to-P architecture in which there is no "smart server, stupid client." With computing power at both the server and client, more flexible, many-to-many computing can emerge.
Geoffrey Smith, president and CEO of one of Consilient's competitors, New York-based Interbind, agrees with the hybrid approach. Interbind also focuses on technology that enables disparate parties to share information and coordinate workflow. With Interbind's platform, for example, an agent bank can use a 2MB client to upload spreadsheets for review with borrowers. Both sides can add information or otherwise tweak the spreadsheets as they work out complex financial deals such as loan syndications.
Interbind's platform is based on eXtensible Markup Language, or XML. Most P-to-P systems rely on the flexibility of XML or Java.
Smith says the Interbind platform reduces lag time and aids process automation because "we don't require people to change the way they do things. Instead of creating some monolithic solution, we say, 'Let's leave the data distributed throughout the systems and run the software where the data is.'"
