X
Business

Microsoft Research has its head in the clouds, too

It's not just the Microsoft product groups that are charging ahead with getting Microsoft "All In" with the cloud. Microsoft Research seems to be stepping up its work on a variety of cloud-related projects, and plans to share details about several of them during the 2010 Usenix Federated Conferences shows in Boston the week of June 21.
Written by Mary Jo Foley, Senior Contributing Editor

It's not just the Microsoft product groups that are charging ahead with getting Microsoft "All In" with the cloud. Microsoft Research seems to be stepping up its work on a variety of cloud-related projects, and plans to share details about several of them during the 2010 Usenix Federated Conferences shows in Boston the week of June 21.

Here are just a few of the new projects from Microsoft Research site that are on the Usenix 2010 docket:

Stout: An adaptive interface to scalable cloud storage

Microsoft Research and UC San Diego researchers

Stout provides distributed congestion control for client requests. Its goal is to improve the performance of applications that are being delivered as scalable multi-tier services deployed in large data centers. These kinds of services often get bogged down with delays and dropped requests under high workloads. Stout is designed to help these kinds of apps adapt to variations in storage-layer performance. "Under heavy workloads, Stout automatically batches the application’s requests together before sending them to the store, resulting in higher throughput and preventing queuing delay," a white paper on the technology explains. The Utility Coprocessor: Massively Parallel Computation from the Coffee Shop Microsoft Research researchers

Nicknamed "UCop," the utility coprocessor is middleware designed to help achieve "dramatic speedups of parallelizable, CPU-bound desktop applications using utility computing clusters in the cloud," according to a white paper on the topic. The project bears some similarity to work done in computational clusters, grids, network file systems and parallel programming, the researchers say. "We devised a scheme that hides the heterogeneity of client configurations, allowing a single cluster to serve virtually everyone," the researchers explain. Interestingly, their prototype is Linux-based. The research paper outlines the design and implementation of UCop in 32-64 nodes running on Amazon's EC2 cloud service. The researchers "introduce the techniques of task-end-to-start consistency and prethrowing to avoid latency penalties" and "avoid bandwidth penalties using a combination of cluster-wide cache sharing, remote differential compression, and the notion of job-end-to-start consistency, the researchers add.

Seawall: Performance Isolation for Cloud Datacenter Networks Microsoft Research and Cornell University researchers

Seawall is about ways to create performance isolation between the tenants co-located on a host in a datacenter. Existing multitenancy schemes, which rely on hypervisors, provide little control over security, VM churn, static bandwidth limitations and compromised hypervisors, the researchers argue. Seawall, "an edge-based solution," is designed to create "fairness across tenant VMs by sending traffic through congestion-controlled, hypervisor-to-hypervisor tunnels," according to a synopsis of the research paper set to be presented as part of next week's Hot Cloud Usenix track.

CloudCmp: Shopping for a Cloud Made Easy Microsoft Research and Duke University researchers

CloudCmp is a new calculation framework designed to help potential cloud-computing customers estimate performance and costs of running legacy applications in the cloud without actually having to deploy them. The framework characterizes the services from competing services, including Google AppEngine, Amazon Web Services and Microsoft's Windows Azure, and then benchmarks the performance and cost of each. The ranking will feature providers who deliver platform-as-a-service offerings, infrastructure-as-a-service offerings, and PaaS/IaaS combination offerings. Distributed Systems Meet Economics: Pricing in the Cloud Microsoft Research, Shanghai Jiao Tong University and Peking University researchers

These Asian-based researchers will be presenting their findings on how cloud computing has "fundamentally changed the landscape of system design and optimization." Their findings, which are focused around Amazon's EC2 cloud service, show how distributed systems affect the economics of pricing. Their findings also are due to be presented as part of the Usenix Hot Cloud conference.

As is the case with all Microsoft Research projects, there's no guarantee as to how/when/if they will become commercialized products. But it's still interesting to see Microsoft putting more of its focus on the cloud in its research organization, and not just the product groups.

Editorial standards