How does the Handle System Scale?
The system scales over two dimensions: • Quantity of requests. If there are typically more requests per unit time than a given server can handle, the load on that server can be reduced by spreading it across more servers at that particular site. If a site becomes overloaded, one or more “secondary” sites may be used to replicate the primary site and thus to offload it. • Quantity of data: If a single namespace becomes too large to be housed on a single server, it can be distributed across multiple servers that make up a site. If a site consists of multiple servers, the handles at that site are evenly distributed among the servers based on a hashing algorithm.
