Papers

Common Service Definition

Jerry Sobieski, Tom Lehman
Mar. 2005
13 pages

Full text: PDF

Abstract

One of the basic obstacles to provisioning end-to-end “light path” services on a global basis is the lack of consistent and measurable performance definitions for these light paths.

Put another way, there is no mechanism to insure that the transport services instantiated by assembling resources through multiple administrative domains and across different network technologies will actually create an end to end path that is predictable, repeatable, and consistent with the user’s request. Fundamentally, there is no clear “service definition” adopted by the network service providers that clearly describes the capabilities and performance characteristics of their services.

This creates a situation whereby the “service” offered or presented at one end of the light path may not be completely satisfied by the network technologies and capabilities across the intermediate networks. These service discrepancies may not be obvious or intentional, or the providers may honestly believe that some minor service differences will not be noticeable or are not significant. Indeed, it may be that the service provided actually exceeds the service requested by the user. But this service variability makes the service non-deterministic, unpredictable, and unrepeatable, and often causes significant performance impairments for the end user/application. There is no objective means of measuring and verifying the service.

In order to enable truly deterministic and consistent network service end-to-end, the users requesting those services and all the networks providing those services must adopt a common service definition. By doing so, the networks providing the services have a standard by which the service can be engineered and constructed, and operationally concatenated and tangibly measured and validated. The users will also have a means by which they can request services that are predictable, deterministic, and again measurable and verifiable.

This paper attempts to define the following:

i) Common Services Paradigm. Define the philosophy, need, and paradigm of a “Common Service Definition” and explain how it can form the basis for global network service provisioning

ii) Common Service Definitions. Define an example common network service. This example service is presented in an appendix and the primary intent is to provide a starting point for community discussion and iteration of specific network service definitions. It is hoped that this can provide a place for multiple common network services to be defined based on various transport services such as Ethernet, SONET/SDH, Fibre Channel, Infiniband, and others.

In general, this paper is addressing the concept of a "service" for the purpose of furthering dynamic allocation of end-to-end data paths with dedicated network resources. In particular, we are looking at “light path” services - services that bear a strong resemblance to traditional connection oriented services or circuits, but which incorporate more recent developments and aspirations for optical and all-photonic networks.

The concept of a “service” is as viewed from an "end-system" which is connected to (and making a request from) a network. In this context, "end-system" is a very general term and may mean a host, computational cluster, data storage cluster, radio telescope, router, or any device which is connected to the network.

As noted, this "service" may traverse multiple administrative domains and network technologies. It is not the intent of this paper to address inter-domain routing, peering, or signaling issues. Likewise, the issue of how a single end-to-end path constructed from multiple heterogeneous network technologies might satisfy a specific service request is not addressed. The service described in this paper is as would be defined, measured, and verified at the end points of the provisioned path. Discussions regarding how to provide these services across multi-domain networks and heterogeneous network technologies are anticipated to be topics for future community discussions. It is expected that the common service definitions will establish an initial set of driving requirements for these other issues.

The remainder of this paper is organized as follows: Section 2 provides a rationale for the need for a Common Network Service Definition. Section 3 provides a framework for a Common Network Service. Section 4 discusses issues with service verification. Section 5 provides an overview of other related issues. Finally, Appendix A provides an example for a basic Ethernet service definition.