Abstract: According to one embodiment, a storage system (400) may include a gateway service (402), metadata service (404) and storage service (406) that each provide different storage system functions. Each service (402, 404 and 406) may include a plurality of servers that are functionally de-coupled from one another. De-coupled servers may service requests without regard to the operation of other servers of the same service. Such a de-coupled arrangement can allow a storage system (400) to be scaled to meet increasing demands, as servers may be added to a system, on a service by service basis, without necessarily re-indexing stored information or reconfiguring multiple instances of a server application. Further, backup servers may be easily added to the system to enable fault tolerance. Servers of a particular service (402, 404 and 406) may be accessed according to a quality of service policy.

Abstract: According to one embodiment, a metadata management system (MDS) may include partitioned migratable metadata. Metadata may be stored in multiple metadata partitions (102-0 to 102-11). Each metadata partition may be assigned to a particular system resource (104-0 to 104-5). According to predetermined policies, such as metadata aging, metadata stored in one metadata partition may be migrated to a different metadata partition. A forwarding object can be placed in the old metadata partition to indicate the new location of the migrated metadata. Metadata partitions (102-0 to 102-11) may be reassigned to different resources, split and/or merged allowing a high degree of scalability, as well as flexibility in meeting storage system needs.

Abstract: A system and method for managing software upgrades in a distributed computing system. The distributed computing system may include a plurality of nodes which provide one or more fault-tolerant services. The system and method perform software upgrades in a sequential or “rolling” manner (e.g., node by node). The rolling upgrade process allows all services and data of the distributed computing system to remain operable and available throughout the upgrade process.

Abstract: A system and method for checkpointing a primary computer process to a backup computer process such that if there is a failure of a primary process, the backup process can takeover without interruption. In addition, upgrades to different version of software or equipment can take place without interruption. The invention provides a lightweight checkpointing method that allows checkpointing of only external requests or messages that change the state of the service instance, thereby reducing the overhead and performance penalties. In addition, the present invention checkpoints data for primary and backups that do not share resources but are logically equivalent. All communication between the primary and backup takes places using network protocols.

Abstract: A system for managing computer resource usage quotas includes a distributed gateway service, a plurality of clients, an internal network, a distributed storage service a system management service, a configuration database, and a distributed metadata service. A method for managing computer resource usage quotas applied to a plurality of users of data includes the steps of defining one or more resource classes, defining at least one usage policy for the resource classes, tracking quota usage for the usage quotas in accordance with the usage policy, and enforcing quota usage for the usage quotas, and storing the data in accord with the resource class definitions. Optionally, the quota usage may be aggregated.