Distributed Services

Corporations increasingly see e-commerce, or collaborative commerce as it is often called, as a strategic tool. According to consulting firm Deloitte and Touche, corporations list the top goals of collaborative commerce as

• Reducing the cost of operations

• Gaining competitive advantage

• Increasing customer loyalty and retention

Collaborative commerce is being built on distributed services. Initially the focus is on Web services; however, complementary P2P technologies might turn out to be the enabler that Web services need to fulfill corporate expectations.

Technology

Distributed P2P services for collaborative commerce fall in three categories; distributed data services, distributed computing, and distributed network services.

Distributed Data Services

A study from IT industry researcher Gartner concluded that large enterprises can create competitive advantages using P2P by developing P2P content networking solutions based on what it calls the "data-centered P2P model." Gartner expects that by 2003 "30% of corporations will have at least experimented with data-centered P2P applications" (http://marketplacena.gartner.com/010022501oth-NextPage.PDF).

Distributed data services move data closer to usage using multiple nodes and sophisticated routing algorithms.

Peer-based storage has distinct advantages:

• Edge devices are plentiful and under-used.

• Users are more mobile, and must access information from multiple locations.

• Centralization has proven costly and prohibitive beyond a certain level of scalability.

• Mobile users are demanding faster access to content.

• Service providers are searching for cheaper solutions.

There are a number of complexities to implementing successful distributed content networks.

Distributed content networks require intelligent caching over a widely dispersed cluster of nodes. Some distributed content networks use predictive seeding to preconfigure the location of data based on usage patterns and known heuristics.

Multisourcing permits a content network to map multiple communication paths to a cluster and/or data store. This is often accomplished through intelligent routing. Security (encryption) is required to ensure integrity of data in transit.

Presence and bandwidth matching guarantees that nodes are available, and that they map performance capabilities appropriately. Content delivery networks automatically detect which peer servers are available on the network to serve the file segments with the greatest overall efficiency.

Distributed Computing

Distributed computing (also referred to as grid computing) attempts to use the idle processing cycles of the PCs on your network. It makes these commodity devices available to work on computationally intensive problems that would otherwise require a supercomputer or workstation/server cluster to solve.

There are usually three fundamental components to the architecture:

• The Network Manager manages client resources, and controls which applications are run on the client machines.

• The Job Manager permits application users to submit work and monitors the progress of this work, and retrieves results.

• The Client manages the running of applications on a client machine.

Distributed computing should continue to gain in popularity as more sophisticated problems surface, such as in genetics and bioinformatics.

Distributed Network Services

The attractiveness of distributed network services lies in their capability to localize traffic, lowering bandwidth expenditures and improving response times. By serving and fulfilling as many requests for data from devices on the LAN, enterprises and ISPs can cut costs and improve performance dramatically.

Bandwidth management technologies give network administrators at Internet service providers or corporate enterprises the capability to set and enforce policies to control network traffic, ensuring that networks deliver predictable performance for mission-critical applications. Bandwidth management systems can prioritize mission-critical traffic, as well as guarantee minimum bandwidth for the most critical, revenue-generating traffic (for example, voice, transaction-based applications).

Webcasting continues to grow, and bandwidth demands continue to increase. Distributed P2P network services automatically scale to use the available bandwidth and computer resources of new participants who request the stream. This is a more economical model for usage in high bandwidth-intensive applications that require infrequent activation.

Load balancing/traffic management products process network traffic streams, switching and otherwise responding to incoming requests by directing these requests to specific server clusters based on a set of predefined rules. These products usually have the capability to test the servers they are connected to for correct operation, and reroute data traffic around a server should it fail. Most of these devices also have the capacity to recognize the requester or the data being requested, and prioritize the request or the response accordingly. The capability of the load balancer/traffic manager to consider the source of the message or the relative load on each of several replicated servers and then direct the message to the most appropriate server increases efficiency and uptime.

The next generation of network services and technologies will continue to offer advantages in caching, load balancing, and bandwidth management. What distinguishes these next-generation services and technologies is their emphasis on further pushing the distribution mantra.

Distributed Networking Products

• Avaki (www.avaki.com) provides a high-level grid protocol that it's adapting to run on top of the XML/SOAP Web Services standard.

• Parabon Computation, Inc. (www.parabon.com) provides grid computing capabilities over the Internet using a Java-based framework.

• United Devices, Inc. (www.ud.com) provides a grid computing platform for corporations and research, along with a global Internet-enabled platform for large-scale distributed computing tasks.

• Porivo Technologies, Inc. (www.porivo.com) provides a service that measures the end-to-end performance of Web applications from the customers' perspective using a network of distributed PCs.

• vTrails, Ltd. (www.vtrails.com) offers a media delivery solution that leverages the power of peer-to-peer networking, smart routing, and edge network capabilities.

• The Cytaq Distributed ResourceNetwork Network (www.cytaq.com) is a network of resource routers linking and integrating varied information resources of an enterprise, such as applications, databases, and so on. The resource routers are small-footprint, Java software services. Multiple information resources can be linked to one resource router and vice versa.