Home TGIF presentations
TGIF 03/23/2012

Migrating Virtual Machines, Quickly, Smartly, and Far Away&em;Advanced Live Migration Technologies for Energy-Efficient and Dependable Cloud Computing, presented by Dr. Takahiro Hirofuchi

Takahiro Hirofuchi, Ph.D. is a researcher of AIST (National Institute of Advanced Industrial Science and Technology) in Japan. He is an expert of system software and operating system and has been working onvirtual machine technologies for five years. Since 2011, he has started a joint work with ACIS Lab of University Florida, regarding disaster recovery and mitigation techniques based on virtual machine migration. In this talk, he introduces advanced live migration technologies developed by AIST, and explains the current status of the joint work. The talk will cover the following topics:

  • Yabusame: A Postcopy-based Live Migration Mechanism for Qemu/KVM
  • xNBD: A Live Storage Migration Mechanism for Virtual Machine Monitors
  • MiyakoDori: A Memory Caching Mechanism Accelerating a Sequence of Live Migrations
  • Kagemusha: A Transparent Mobile IPv6 Tunneling Mechanism for Wide-Area Live Migrations
  • NSF/JST RAPID Joint Project - IT Virtualization for Disaster Mitigation and Recovery


TGIF 03/16/2012

MatchTree: Scalable and fault-tolerant resource discovery in WAN using distributed matchmaking and aggregation, presented by Kyungyong Lee

MatchTree, a wide-area resource discovery framework which builds upon a Peer-to-Peer (P2P) framework to deliver scalable and fault-tolerant service supporting distributed query processing and aggregation of results. MatchTree leverages a self-organizing tree for query distribution and result aggregation with the latency cost scales with O(log N), where N is the number of queried nodes. This presentation will discuss the overall architecture of MatchTree and propose heuristics to improve fault-tolerance and reduce query response times through redundant query topologies, dynamic timeout policies, and sub-region queries. Additionally we will show qualitative evaluations of MatchTree through a large scale simulation, as well as experiments with a prototype MatchTree implementation deployed on a wide-area infrastructure (PlanetLab).

TGIF 02/24/2012

SOLARE: Self-Organizing Latency-Aware Resource Ensemble, presented by Heungsik Eom

This paper proposes and evaluates Self-Organizing Latency-Aware Resource Ensemble ( SOLARE), a self-organizing and managing cluster systems based upon utility functions. In contrast to previous works, SOLARE is a fully decentralized clustering algorithm without any central units such as servers, super peers, cluster heads or landmarks so that it is robust against the single point of failure. Furthermore, SOLARE provides the adaptability on dynamic network conditions by monitoring the utility of cluster and migrating into another high utility valued cluster. Our evaluation shows that SOLARE is able to guarantee satisfying user demands expressed by system parameters in terms of intra cluster latencies and the number of cluster members. Also, we verify the ability of SOLARE to adapt on dynamic network conditions by showing that the percentage of the number of nodes which migrate into another cluster and average utility value as nodes join SOLARE.

TGIF 02/10/2012

Litter: A Lightweight Peer-to-Peer Microblogging Service, presented by Pierre St. Juste

Microblogging has become an important part of the social web evolution and is being utilized in many aspects such as advertising, political campaigns, and popular uprisings. Due to its heavy centralization, many have proposed decentralized alternatives based on a variety of models. This paper suggests a fully distributed approach built on top of existing peer-to-peer technologies. We demonstrate that, by exploiting the services of current peer-to-peer middleware along with the properties of the social graph, it is possible to create a simple, yet practical microblogging service that is impervious to many of the shortcomings of their centralized counterparts. The approach has been implemented as a software prototype that is readily available for download in order to test our design in real life environments.

TGIF 03/18/2011

Tag Management Architecture for Hardware-Based Translation Lookaside Buffers in Virtualized Platforms, by Girish Venkatasubramanian

The use of virtualization to effectively harness the power of multi-core processors has emerged as a viable solution to meet the growing demand for computing resources, especially in the server segment of the computing industry. However, one significant issue in using virtualization is the reduction in the performance of virtualized workloads, compared to the non-virtualized counterparts, caused by the hardware-managed Translation Lookaside Buffers (TLBs). This talk proposes the tagging of the entries of hardware-managed TLBs with process-specific identifiers using the Tag Manager Table (TMT). The design and working of the TMT are discussed. A full-system execution-driven simulation framework for investigating the performance of TLBs on virtualized platforms is developed. Using this simulator, the benefit of the TMT, in terms of the improvement in TLB miss rate and workload performance, is investigated using the simulation framework and the various factors that affect this benefit of using the TMT are analyzed.

<< Start < Prev 1 2 3 4 Next > End >>

Page 1 of 4