Caching Protocols Move to Cloud to Make High Performance Computing Cheaper
15 December 2011
A new cloud cache system using well known caching protocols gives users with large data requirements the computational grunt they need without having to invest in more hardware.
Called Cache as a Service (CaaS), the cloud service model has been developed at the University of Sydney. It will make high performance computing services more readily deployable onto the cloud for scientists - including physicists and astronomers - and industries like finance and e-commerce, which have massive Input/Output (I/O) requirements.
Professor Albert Zomaya from the School of Information Technologies says CaaS allows users to more "cost efficiently" deploy applications with heavy I/O activities on the cloud.
"CaaS incorporates a pool of remote memory and makes it available transparently to users as a service additional to typical cloud services (i.e. infrastructure as a service or IaaS)," he says. "The elastic cache system developed as part of CaaS enables users to dynamically rent the memory needed for your computing requirements.
"This much helps I/O-intensive applications like e-commerce websites run faster with minimal extra cost. A CaaS offering lets you pay for your requirements as you go, without incurring extra overheads.
Professor Zomaya says CaaS combines a system architecture level mechanism and service level approach to better deal with I/O-intensive user applications in the cloud without any modifications to existing applications: "We use the same programming protocols you would find in traditional caches, so there are no hidden overheads."
Since extra cache memory in CaaS is provided at the operating system level it is not visible and accessible by users other than its owner; hence, security is not a concern.
CaaS was developed by the University's Centre for Distributed and High Performance Computing, directed by Professor Zomaya, in conjunction with the Distributed Computing Systems Lab, in Seoul National University. The system is currently being peer reviewed but Professor Zomaya says it is ready for use by cloud providers or organisations with private cloud systems.
Contact: Jocelyn Prasad
Phone: 02 91141382