Cloud Physics, Inc.

Nohhyun Park

Carl Alan Waldspurger

Alexander T Garthwaite

Irfan Ahmad

Contention for a resource in a computer system resource is managed by measuring a resource performance metric and, for each of a selected plurality of clients (for example, virtual machines), a client performance metric. For each of the selected clients, a relationship measure, such as correlation, is determined as a function of the resource performance metric and the respective client performance metric. A degree of resource contention effect is determined for each of the selected clients as a function of the respective relationship measure, and a resource-related action is taken according to the respective relationship measures. Clients may include virtualized components contending for storage. Example metrics include functions of I/O operation counts, latency or throughput measurements, pending I/O request counts, I/O throughput relative to I/O latency, a degree of change of the respective clients' I/O behavior, etc. Possible actions include changing resource allocations, access throttling, and reporting, etc.

Relationship-based resource-contention analysis system and method

The configuration of a cache is adjusted within a computer system that includes at least one entity that submits a stream of references, each reference corresponding to a location identifier corresponding to data storage locations in a storage system. The reference stream is spatially sampled using reference hashing. Cache utility values are determined for each of a plurality of caching simulations and an optimal configuration is selected based on the results of the simulations.

Dynamic cache allocation adjustment using multiple cache simulations for efficient cache utility curve construction

Interaction is evaluated between a computer system cache and at least one entity that submits a stream of references corresponding to location identifiers of data storage locations. The reference stream is spatially sampled by comparing a hash value of each reference with a threshold value and selecting only those references whose hash value meets a selection criterion. Cache utility values are then compiled for those references. In some embodiments, the compiled cache values may then be corrected for accuracy as a function of statistics of those location identifiers over the entire stream of references and of the sampled references whose hash values satisfied the selection criterion. Alternatively, a plurality of caching configurations is selected and the selected references are applied as inputs to a plurality of caching simulations, each corresponding to a different caching configuration. A resulting set of cache utility values is then computed for each caching simulation.

System and method for efficient cache utility curve construction and cache allocation

Cache utility curves are determined for different software entities depending on how frequently their storage access requests lead to cache hits or cache misses. Although possible, not all access requests need be tested, but rather only a sampled subset, determined by whether a hash value of each current storage location identifier (such as an address or block number) meets one or more sampling criteria. The sampling rate is adaptively changed so as to hold the number of location identifiers needed to be stored to compute the cache utility curves to within a set maximum limit.

Cache allocation system and method using a sampled cache utility curve in constant space

Cache utility curves are determined for different software entities depending on how frequently their storage access requests lead to cache hits or cache misses. Although possible, not all access requests need be tested, but rather only a subset, determined by whether a hash value of each current storage location identifier (such as an address or block number) meets one or more sampling criteria.

Growing community of inventors

Redwood City, CA, United States of America

Nohhyun Park