FAQ Sign-up Sales Support Billing Contact Us Site Map
 

Subscribe to Email List

Press Room

Download this Press Coverage (94.91kb)
May 20, 2002
Computer Letter: The Price Is Right
 

An intriguing startup called InvisibleHand Networks wants to bring Adam Smith’s famous metaphor to life all over the Net.

Several years ago, the business-to-business bug bit big and everyone, it seemed, was talking digital marketplace. Rivers of venture money flowed, hundreds of startups popped up, and B2B conferences spilled over as eager entrepreneurs’ eyes danced with visions of building a whole new parallel universe. It felt almost as if mankind had finally discovered the Internet’s true raison d’etre: to serve as the freest of free markets where every buyer and seller, from every spot on the globe, could come together and do business as never before.    

Maybe B2B fever was a millennial phenomenon, a momentary but inevitable plunge into irrationality that occurs every time the calendar shows an excess of zeros. Or, it may have been the logical extension of so much free-market policy-making during the ‘80s and ‘90s. As it turned out, of course, the “frictionless commerce” trumpeted so loudly by Bill Gates and others was but a daydream-unless you count the near-total lack of traction achieved by most would-be marketplace operators.     

More reasonable an explanation is the fact that while the Internet truly does have the potential to create entirely new kinds of markets, the technology required just wasn’t ready. Markets, like the village and towns that traditionally have grown up around them, are exceedingly complex entities, perhaps describable only through metaphor and perhaps only one aspect, or layer, at a time. Is Chicago a set of streets, the buildings on those streets, the people in those buildings, or the activities in which all those people engage? All of that and more, we’d say. And the Chicago Mercantile Exchange? Is that just the famous “pit” where traders jostle each other to make deals? Or is it the prices and other numbers that their shouts, nods, and raised fingers signify? Again, that an more, including all the bodies, minds, strategies, prices, 1s and 0s flying through computers, and even obtuse abstractions like 90-day pork-belly futures.

Under the hammer
Getting a computer to faithfully simulate such an incredibly rich but quite unchoreographed dance turns out to be about as tricky a programming challenge as has ever been considered. So, it’s no wonder that virtually all grand B2B schemes fell flat on their face, unrealized beyond an overheated press release or two. And it’s no wonder, either, that the few successful B2B startups there are have each been devoting themselves to building essentially one aspect or function of the electronic marketplace: catalog management, strategic sourcing, product information, and so forth.    

One of the more compelling concepts to emerge from all this activity was that of dynamic pricing. With buyers and sellers interacting electronically and near-instantaneously, the theory goes, prices can be permitted to rise and fall according to the laws of supply and demand, profits may be maximized, and commodities allocated in the most efficient way possible. Thus, a growing number of companies were seen setting up private auctions to obtain better prices from their roster of suppliers while other, including even IBM, began selling finished goods through their own auctions and those run by eBay and others.

Agent agenda
Of all the outfits we’ve encountered working with dynamic pricing model, one of the most intriguing is InvisibleHand Networks. It’s focused on one of the more vital but intangible commodities being sold online these days, namely IP bandwidth and IP-based services. The firm has designed its software to help internet service providers create what amounts to a spot market where they can sell their excess bandwidth in ways-and at prices-that until now have been impossible. The software, called Merkato, relies on a set of distributed agents and proprietary algorithms-base on game theory and economic theory- that create something larger, in a sense, than the sum of their parts.

The wasteland
As it is, ISPs have generally employed circuit-based business models, which tend to leave affair amount of potential revenues on the table. ISPs sell their available bandwidth to customers as fixed-price packages, or circuits, under fixed-period contracts. They price these packages according to what they figure the market will bear-say, $500 a month for a T1-capacity line. Based on the fact that most Net traffic is bursty in nature-downloading a Web page consumes only a momentary spike of packets and e-mail takes up little capacity at all-the ISP typically reckons that on average, each customer will actually us one 40% of the rated capacity of his “circuit.” (In effect, the customer pays for the ability to have immediate access to the circuit’s maximum capacity during those times when he needs it.)    

This was of building slack into the pricing has mixed results. On the one hand, it enables the ISP to oversell its “circuits,” with their aggregate maximum capacity exceeding what the ISP would actually be able to provide. On the other hand, if too many customers consume their full share of capacity at the same moment, each may experience degraded service. Worse for the service provider, most of the time a good deal of capacity goes unused and essentially wasted. Were the right mechanisms in place, this slack could be sold at a profitable price.

Big squeeze
Company officials explain that he Merkato software eliminates what amounts to a supply-based guessing game with real-time, dynamic, demand-based pricing. This approach, they claim, can bring ISPs extra, high margin revenue by squeezing a major inefficiency out of their business model.

Risky bid-ness
Merkato recognizes IP bandwidth as the unique commodity it is. It is shareable, as unused capacity can be made available to others. Indeed, consumption of bandwidth can take place at the moment it is purchased; as with gas and electricity, there is nor waiting for delivery. But it is perishable, too. It cannot be stored for later use but must be consumed now. Demand for IP bandwidth may fluctuate rapidly over time imagine the momentary stress of a video conference, for instance. And finally, it shows strong price elasticity: it’s relatively easy for customers to switch between suppliers and, given the right incentive, they will.     

Given all this, the trick for ISPs is to come up with the right price at which to sell the excess capacity they have available at any given moment. Merkato helps by creating a spot marketplace that discovers and negotiates that price based on inputs from the seller and all potential buyers. Once a customer has bought a chunk of bandwidth, the Merkato system is able to automatically communicate with the appropriate networking gear-usually an edge router attached to the seller’s net- and get the bandwidth provisioned to that customer. In a typical setup, the system undertakes this bidding and price-discovery process every 5 minutes.

Emergent behavior
The company emphasizes the radically distributed nature of its software. Each of sellers and buyers in a Merkato marketplace are represented within the system by their own autonomous agent, programmed to pursue their specific goals regarding, quality of service (QoS), and other parameters. From the complex interaction of these independent an “intelligent” agents emerges, in theory, the most efficient and mutually beneficial allocation of resources possible-in effect, the invisible hand classical economist Adam Smith wrote about some 200 years ago.    

We’re impressed by the far-reaching vision behind Merkato. The system is designed so that real-time data bout traffic levels can continuously be fed back into each buyer agent, thereby enhancing its ability to make correct decisions about what to buy and for how much.

Get the message
The Merkato exchange can be set up to serve in several different locations on the Net.  The simplest would be at a so-called peering, or co-location, point, where numerous networks link together to exchange traffic. Buyers would be enterprises with traffic to send over these networks, content providers with servers located at the peering point, or ISPs buying bandwidth from peer networks.  In addition to creating spot markets, the system makes possible reservation markets. There, buyers could reserve a certain amount of capacity for a pre-arranged price-in effect, a hedging strategy.     

InvisibleHand tells us that one of its main achievements has been to make its software scale well to a very large number of diverse and self-interested participants, each represented by its own agent. Done the wrong way, so many agents would require too much computational horsepower and internal message processing to be practical. But the Merkato system is built around a core of shared market rules and other resources and a lightweight signaling scheme. Depending on the size of a Merkato marketplace and its volume of activity, it can be run on a single machine or distributed across multiple platforms each run independently of the others.

Balancing act
The decentralized architecture, with a virtual “micro-market” designated for each type or level of service being offered, enables sellers to offer just the right mix of those services. History has shown, InvisibleHand tells us, that most new types of Internet traffic-streaming media, voiceover-IP, and so on-have emerged from out of the blue, catching ISPs pretty much by surprise. They’ve often over-or underestimated demand for each type. And except for over-provisioning, they’ve not had a way of making sure demand for one traffic type doesn’t interfere with the others. The company’s value proposition is simply stated and quite compelling, we believe: With billions of dollars invested in IP infrastructure and budges as tight as a drum, telecommunications providers need ways of squeezing more profits form that investment. Over time, value-added services will have their day, but without a way to bill more efficiently for existing capacity, even those services will not deliver the returns carriers and ISPs need to show their investors. InvisibleHand says it setup can increase revenues by 45%, boost gross margins by as much as five times, and still help buyers enjoy 30% savings on bandwidth.

Liquidity
Already, we’ve seen numerous attempts at creating bandwidth exchanges, run by firms such as Arbinet, Band-X, and LighTrade (recently defunct.) They’ve focused mainly on long-distance telephone calls, enabling telcos to buy and sell minutes in short-and long-term contracts. (If it isn’t going on already, the time is near when such exchanges begin to influence the routing of individual telephone calls.) Now, IP transit, available in massive quantity, easily provisionable, and deliverable at any level of QoS, looks like a perfect trading opportunity, too- if only pricing issues could be worked out. IP bandwidth trading is just starting to show up as a business and InvisibleHand tells us that it views all players in that sector as potential customers for its software.     

Clearly, right now, any discussion of trading IP bandwidth takes place wit the Enron house-of-cards debacle still reverberating in everyone’s ears. Indeed, InvisibleHand tells us that Enron was a Merkato beta customer, trying out the software in support of its own (now highly suspect) bandwidth-trading activities. But the startup points out that Merkato is a product available for all carries and providers to use, while Enron’s game was to offer trading services and, as monkey in the middle, collect a percentage for itself.    

InvisibleHand’s flagship customer is Telehouse America, partly owned by Japan’s KDD, which operates a handful of collocation and neutral peering facilities around the country. Telehouse is using the technology to make its peering more liquid: customers with services or content to deliver can buy bandwidth as they need it, on a Merkato-based spot market. Meanwhile, InvisibleHand has set up StreamingHand, a service that supplies bandwidth, priced by Merkato, to some 65 streaming content providers. For these companies, bandwidth accounts for as much as 90% of their cost-of-goods-sold. StreamingHand is expected to reach cash flow break-even this quarter.

Hoppity, hop, hop
InvisibleHand’s business model is licensing-based but suitably flexible. It offers to collect transaction-based fees or charge for each participant in a Merkato-based market. In short, as customers gain value from the software, so does the startup.     

The company’s technology stems from research originally undertaken at Bell Labs and Columbia University. Much of the thinking has gone into basic market mechanisms, including something called the progressive second price auction algorithm. It is described as allocating “variable-size shares of a resource among multiple users” in a way that maximizes total user value. Equally important is that Merkato is designed to handle any network topology or business model a customer wishes to pursue. For instance, the company has described in one of its whitepapers a hop-by-hop model, I which buyers would “construct their own paths or virtual networks by buying form a large number of [Merkato-based] micro-markets, one for each link in a network.”

Alas, poor network
As might be expected, this work is part of a larger research trend into online markets, auctions, and even methods for getting competing agents to cooperate to mutual benefit. Evidently, markets are widely envisioned as a good way of allocating and pricing all sorts of electronic services that show much the same fluidity as IP bandwidth: content delivery, server-grid-based computation, and network storage, for example. We can imagine a future, too, where wireless networks might sell radio bandwidth to mobile virtual network operators (MVNOs) and others on spot markets, especially as wireless data services come online.    

A quick delve into Google reveals that research into market-based bandwidth allocation is underway at numerous companies, IBM, Sun Microsystems, Fujitsu, and British Telecom (BT) among them. Because of its shared, decentralized nature, IP bandwidth and services present some fascinating technical challenges related to pricing and allocation. Always lurking is the so-called tragedy-of-the-commons: If one too many self-interested users hog a shared resource, he can bring it down to the point where it’s no longer available to himself or anyone else.

Traffic jam
Hewlett-Packard, BT and Norwegian PTT Telenor, along with several Europeans universities, have been working on something called the market-managed multi-service (M3I) Internet project. Its premise, to quote the group’s website, is “that a simple packet network may be able to support an arbitrarily differentiated set of services by conveying information on congestion from the network to intelligent end-nodes, which themselves determine what should be their demands o the packet network. There would then be no need for large buffers or priority queues within the network, or connection acceptance control at the border of the network.”  A key technology t make such a scheme work is a proposed addition to TCP called explicit congestion notification (ECN), which helps feed information about congestion back to the origination point of a transmission. There, decisions can be made about alternate routing.     

If there’s anything that worries us about InvisibleHand’s business it’s that we’re not entirely sure the marketplace is ready for such a sophisticated product. It may turn out that many ISPs and other service providers customers are happy to go along with today’s way of doing business, paying a little extra, perhaps, but having a fairly straightforward set of purchasing decisions to make InvisibleHand will therefore have to do its best to educate the market. No doubt, showing them hard dollar savings will be the best way to convince service providers that, as Karl Marx stated after pondering Adam Smith’s handiwork, there’s nothing to lose but a few chains.

 

Volume 18, Number 13

May 20, 2002

 
Back