System to Improve Cloud Service Datacenter Performance – IP 1530

Distributed computing architectures widely useCloud service datacenters to meet the performance of servers as many computingapplications are adopting a “cloud service” computing model. Typical examplesof such architectures include physical or logical separation of the computationdomain and the storage domain or the usage of distributed replicated storage,where a host server replicates its data to remote servers. With these newarchitectures, many high-performance servers in the datacenter either locallycache, or store a copy of the data in a distributed storage system or, in thecase of a centralized storage system, locally cache the remotely-storedpersistent data. Such caching is performed in order to improve data accessperformance. When an application using the cloud service needs to update thedata being used from a local cache or store, challenges can arise due to datatransfer latencies between the layers of a distributed architecture. FIU inventors have developed techniques andsystems for enabling local independent failure domains in a host server or datacenter to address the challenges using a communications protocol between theLA-IFD and its host server. These techniques include receiving a request towrite a data segment to persistent storage; synchronously storing the datasegment in a buffered data segment at the LA-IFD and initiating an asynchronousupdate of the data segment at a remote storage system; sending a writeacknowledgement indicating completion to the requestor; and, after receiving acompletion notification from the remote storage system, removing the buffereddata segment from the LA-IFD. In some cases, techniques allow a host server andLA-IFD pair to monitor one another for failures and implement a modifiedprotocol in the event of unavailability. Shantanu Balkundi sbalkund@fiu.edu 305-348-8061