 |
Welcome to the FWGrid Project
Introduction
FWGrid is a research infrastructure project in the UCSD Dept. of Computer Science and Engineering, and was sponsored by National Science Foundation from 2003-2011, Award # 0303622 ("RI: FWGrid: A Research Infrastructure with Fast Wireless, Wired, Compute, and Data Infrastructure for Next Generation Systems and Applications"). It continues to provide resources to researchers. It was an outgrowth of the NSF-sponsored UCSD Active Web Project.
FWGrid has enabled exploration with real systems and users, radical new applications, novel application structures, system architecture, resource management policies, innovative algorithms, as well as system management. It fundamentally enhanced many activities of the department. FWGrid modeled a world where widespread high bandwidth wireless, extreme wired (fiber) bandwidth, and plentiful wired computing and data resource are the norm. FWGrid was used to support a wide variety of research, ranging from low-level network measurement and analysis, to grid middleware and modeling, to application-oriented middleware and new distributed application architecture and finally to higher level applications using rich image and video, both off and on line.
Equipment Infrastructure
The FWGrid digital fabric is characterized by computing elements that are distinguished by their high networked bandwidth, powerful compute and data capabilities, and richly diverse input/output capabilities (e.g. graphical displays, cameras, GPS, laser range-finding, etc.). These elements are knit into large scale distributed applications and resource pools known as a computational grid. FWGrid has enabled research in innovative and radically new applications, systems and system architectures, and the emerging technical and even social use of systems and services built for this digital fabric.
Key aspects of this infrastructure included:
-
Distributed clusters with large compute (teraflops) and data (10's of terabytes) capabilities (to power the infrastructure)
-
Rich wired networks of 10-100Gbps (move and aggregate massive data and computation)
-
High-resolution large-screen displays distributed throughout the CSE department
FWGrid has had a transformative impact on the department, accelerating our growth into experimental computer science, continuing the transform and broaden both undergraduate and graduate education, providing a modern Grid infrastructure on which to perform research experiments, but one that also extends to wireless and peer-to-peer computing/storage. Because of the fortuitous timing with respect to our new CSE building, the infrastructure has had a deep impact on the department's educational, research, and social environment. It has supported experimental research, multi-faculty and multi-disciplinary collaborative research and education. FWGrid has formed a nexus for systems research within the department, and coupled all of us to researchers on the campus, to SDSC and CalIT2, the metropolitan area, and wide area at high bandwidth. This has enabled innumerable shared experiments and collaboration. In short, it has coupled us to the future at high speed.
Our equipment, which continues to be operational, includes a multi-level high bandwidth Ethernet network infrastructure (gigabit aggregated to 10Gigabit) that provides a full 1Gbps bisection for each node. The system is a number of large clusters of dual Opteron & Xeon systems, providing scalable compute and storage capabilities. Cluster configurations include 96 HP DL145 dual-Opteron 1.6GHz servers with 2GB RAM, two HP DL145 dual-Opteron 2.2GHz servers with 16GB RAM, 64 Dell SC1425 dual-Xeon 2.8GHz servers with 4GB RAM, and 160 Dell SC1425 dual-Xeon 3.2GHz servers with 4GB RAM. Each of the servers have two 250GB hard disks, providing and aggregate cluster storage capacity of nearly 160 terabytes. We have dedicated one of the DL145 1.6GHz servers as a computing frontend, an access portal. It monitors and maintains the cluster resources with SGE job scheduler and the OS images for cluster nodes. Another DL145 1.6GHz server is responsible for user administration; managing user credentials as well as hosting all user home directories. The remaining 318 servers make up the computing cluster which provides, aggregately, at least 1.5 Tflops CPU power, ~1TB RAM and ~160TB of disk storage.
All of the compute servers have two built-in 1Gb network interfaces. The 320-node cluster is divided into 10 rack groups (32 nodes/rack). The network connectivity within a rack is provided by either an Extreme Networks Summit 400 switch or an HP 3400cl (48 Gigabit Ethernet ports and 2 10Gb uplinks). The network connections amongst the first 5 racks are provided by Fujitsu XG800 (12-port 10Gb L2 Ethernet switch). The last 5 racks are connected by HP 5406zl (12 port 10Gb L2 Ethernet switch). This configuration provides a research network with 26Gb bisection bandwidth. All the computing resources (cluster front end, cluster computing nodes and the 64-bit computing servers) have direct network connectivity to the public Internet (1Gb shared bandwidth).
The FWGrid infrastructure also includes a wealth of computing elements, ranging from small video cameras to rich fixed and head-mounted displays to clustered 64-bit fast compute engines to 10's of terabytes of storage (100x current infrastructure), from wireless clients with 100-500Mbps (100x current practice) and wired network infrastructures with 10-100Gbps switched connections (100x current practice). The wired networks' massive bandwidth allow the fixed resources (computing and data bricks) to be dedicated to local activities, clustered or even aggregated in units of any size up to the entire infrastructure (and with resources outside the building). The fast wireless networks have been implemented by using commercial products configured to use multiple channels of the allocated frequency band simultaneously, leveraging commercial products and achieving multiples of the current nominal bandwidth. Such high wireless bandwidth have enabled mobile clients to stream/receive multiple high resolution video or data streams, enabling rich interaction with the environment, based on computing and data in the core.
During the last few years, we completed the installation of large-screen public display resources in CSE. This was motivated by the fact that while FWGrid was scaling up the problems that could be solved, there was no easily accessible resource for viewing the large-scale outputs. Additionally, it is hoped that other projects, not necessarily data- or compute-intensive, could leverage these displays for their research.
While the FWGrid Project is now officially over, the CSE Dept. is seeking ways of continuing the availability of its resources for future researchers.
|
|
