Check out the new USENIX Web site. next up previous
Next: Extensibility Up: Active Name Architecture Previous: Active Name Architecture

Overview


  
Figure 1: Active Names Architecture Summary.
\begin{figure}

\begin{center}


\scalebox {.5}{\includegraphics{arch3.eps}}

\end{center}
\small\em
\end{figure}

A client that wishes to access a service constructs an Active Name for that service consisting of a name to resolve and the name of a namespace program to resolve it. The client then hands the name to the nearest resolver, which executes the namespace program to begin resolution of the name. Figure 1 illustrates the basic Active Names architecture.

A namespace program has two tasks: it must locate the next program to run and then transport data to that program. In doing so, namespace programs effectively establish a path through a series of namespace programs from request-source to reply-sink. Each program then acts as a filter that transports and transforms its input to its output.

A program locates the next program to run using two mechanisms. First, each Active Name is resolved recursively using a hierarchical delegation mechanism: a client specifies a root namespace program which partially resolves the name and determines which namespace program has jurisdiction over the remainder of the name; the root program then hands the partially-resolved name to the next namespace program, which continues the process. Second, to support composibility of services and to increase efficiency, the programming model follows a continuation passing style: as names are resolved the system constructs an after-methods list that uses a list of Active Names to describe a pipeline of services that will transport a result to its destination. Thus, once a name is recursively resolved to a service, rather than returning the service's output through the same call path used to resolve it, the leaf namespace program pops the top Active Name from the after-methods list and resolves that Active Name and the remaining after-methods list to transport the service's output to the client.

Resolvers provide the basic resource necessary to execute namespace programs, and they are distributed through the system. In principle, resolvers may be located anywhere, but in practice they are most useful when they are located at "interesting" points in the network: near clients, near servers, or near bottleneck network links. We envision a system that provides suitable resolver infrastructure at such points.


next up previous
Next: Extensibility Up: Active Name Architecture Previous: Active Name Architecture
Amin Vahdat
8/31/1999