SLASH2 Testing in a Wide Area Environment

Project Information

Discipline
Please Select... 
Orientation
Research 
Abstract

SLASH2 is a file system developed by the PSC that supports multiple data residencies through user managed replication. The purpose of this project is to research the performance of the file system across the wide area network and on different platforms. Tests will be run between Future Grid resources that are geographically separate, local Future Grid resources, and resources at the PSC. Tests will include 1) the I/O performance of clients accessing local and remote resources, 2)server to server data replication performance, and 3) performance of distributed metadata function in normal operation as well as disrupted network situations. More information about SLASH2 is available at : http://quipu.psc.teragrid.org/slash2/

Intellectual Merit

The Intellectual Merit of the project is the new file system paradigm that removes the data replication burden from the user and incorporates that task in system-level operations.

Broader Impacts

SLASH2 is a new type of information tool that will enhance the infrastructure for research. By allowing easy access to data at multiple locations, researchers will be able to spend more time on science and less on routine data management.

Project Contact

Project Lead
J Ray Scott (J Ray Scott) 
Project Manager
Zhihui Zhang (zhihui) 
Project Members
Nathan Stone, Paul Nowoczynski, Jared Yanovich, Robert Budden  

Resource Requirements

Hardware Systems
  • Network Impairment Device
  • I don't care (what I really need is a software environment and I don't care where it runs)
 
Use of FutureGrid

We will install SLASH2 i/o daemons (sliod) on two systems within the FutureGrid. Ideally these would be systems at one site. Later we would add remote sites to broaden the tests. In addition to the sliods, we will deploy one metadata server process that will be the core of the test file system. All processes are user level. Ideally the processes will persist on the hosts. We are particularly interested in using the network impairment device to introduce network problems and see how the file system performs and recovers.

Scale of Use

We would like to use a few terabytes of data on each system we have access to. We will move the data from system to system using the SLASH2 replication. We will want to monitor system and network performance. After the SLASH2 file system is in place, we will run standard I/O benchmark tests such as IOR, bonnie, etc.

Project Timeline

Submitted
08/24/2011 - 22:04 
»