A multi-domain server farm is a collection of servers divided into a number of distinct domains. While the resources of the server farm are shared, each domain is isolated from the others. Server farm administrative software allocates resources, and enforces isolation policies. This paper describes GulfStream, a distributed software system that addresses the problem of managing the network topology of such a server farm. In particular, it addresses the following core problems: topology discovery and verification, and failure detection.

Unlike most topology discovery and failure detection systems which focus on the nodes in a cluster, GulfStream logically organizes all of the network adapters of the server farm into groups. Each group contains those adapters that can directly exchange layer 2 messages. GulfStream dynamically establishes a hierarchy for reporting network topology and availability of network adapters.

We describe a prototype implementation of GulfStream on a 55 node heterogeneous server farm interconnected using switched fast ethernet. This was done in the context of the Oceano project which provides for dynamic resource allocation in response to workload variations. We discuss scaling GulfStream to larger environments; we envision its use in server farms containing thousands of nodes.