Project: HPN2000 - High Performance Next-Generation Networking Environment for research into high performance computing/networking

FINAL REPORT

Principal Investigator (Simon Fraser University): Peter Anderson

Collaborating Investigators (Simon Fraser University): Loki Jorgenson, Stephen Braham

Funding: $12K from C3/HPCnet, $5K from Telematics, $7K from CECM/PDG.

Hiring: Full-time development person, 6 months. Extra resources from senior management team at PDG, and experts at Telematics.

Project HPN2000

Proposed Deliverables

• Upgraded disaster information resources
• Integrated VEMIS/PolyLAB systems
• Java-based status and monitoring interface

Summary

The supported project was aimed at porting Telematics emergency preparedness/disaster response systems over to PolyLAB systems, and then integrating them into a powerful resource, using a combination of HPC systems and advanced networking.

Project Management: PDG and Telematics Research Lab co-managed the project.

Primary Task Completion Description

The level 2 Work Breakdown Schedule (WBS) is detailed below.

1.1: Training of Experts

PDG and Telematics have experts in various technologies that are applicable to 21^st Century networking. Cross-training will take place to produce team members who are fully-versed in all systems.

Results

The cooperation between the two groups has produced an amazing amount of cross-training, and has led to new projects that utilize the talents of the combined technical teams developed during project execution. James Nedila was hired as a new junior programmer in the PDG. He was then trained in Telematics systems, learning radio communications, layout of disaster preparedness information systems. Peter Anderson and Gordon Gow, of Telematics, were introduced to ATM technology, multicast IP protocols, and large-scale computing. The entire programming team of the PDG was introduced to satellite communications technology, and has proceeded to develop a specialized understanding of integrated networking technology. The newly developed skills have been applied to disaster communications, and have been requested by NASA for manned Mars exploration mission technology development.

1.2: Transfer of Telematics web resources to PolyLAB systems

Telematics disaster information resources will be ported to the PolyLAB HPC-class delivery system (Kasei), and the backup development system (Nirgal). Kasei will host web services via multiple host names. Kasei and Nirgal will supply disaster preparedness systems DNS service.

Deliverable: A range of emergency and disaster preparedness information sites with a range of updated web tools.

Results

A wide range of services now reside on Kasei. These include:

• Emergency Preparedness Information Exchange (EPIX)

• United Nations International Decade Natural Disaster Reduction information services

• Emergency Preparedness Canada

• Safe Guard (An emergency preparedness public awareness site)

• BC Provincial Emergency Program

• BC Inter-Agency Emergency Preparedness Council

• BC Emergency Social Service Association

• Industry Canada Emergency Telecommunications

• Hazard Net (Telematics led international demonstration project)

Figure 1. Enterprise 450 Server

A second site has also been developed with the NATO Civil Protection Committee, using much of the technology developed for the above, including XML/OpenMath encoding of disaster response information to support natural disaster mutual aid efforts among nations in Europe and North America.

Kasei is an Enterprise 450 server, with 4 redundant processors, 2 redundant internal power supplies, optical ATM and 100 Mbps networking, and large amounts of disk space and memory. The Kasei system has been placed in a secure area, with UPS backup and emergency power provisions. The room in which it is located has a halon fire-prevention system.

1.3: Construction of integrated network system

Communication systems will be integrated. Satellite communication systems will be tested, and ATM connectivity established. VEMIS wireless network will integrated into global high-speed networks.

Deliverable: Test results from ATM over space-based link communication. Integrated wireless networking/space-based link/conventional internet system. Graphical network status and control tools.

The range of integrated networking solutions implemented has received a large amount of national and international attention. A large, flexible, Integrated Network was built, augmenting the Virtual Emergency Management Information System (VEMIS).

Figure 2. The VEMIS Integrated Network.

Working with Newbridge Ltd, a MainStreet ATM switch was integrated with the satellite communication system. TCP/IP packets were then encoded into ATM packets by the switch, using the CLIP standard, transmitted via T1 to the satellite modem, and uplinked via the Anik E1 spacecraft to the Communications Research Centre in Ottawa. At standard operating levels (6 dB signal to noise ratios), there were no observed packet losses. Similar VSAT technology will be used by Telematics, PDG, and NASA to transfer data from the Haughton Crater, Devon Island, in Nunavut, back to NASA Ames Research Center. This work will happen as part of a Mars analog test of manned mission technology, this Summer.

Figure 3. The VSAT satellite dish at SFU, pointing towards the Anik E1 spacecraft. Data is relayed from the Integrated Network server to Geosynchronous Earth Orbit, and back down to remote locations on the planet.

The wireless network, based on 56 Kbps packet radio technology, was used for a variety of tests, fully integrating it with the information services located on Kasei. It provided network services in the Thunderbird IV tests, a joint Provincial/Federal emergency exercise. The radio system is based on a set of repeater stations located in the Greater Vancouver Regional District. Remote sites use the repeaters to broadcast signals back to internet gateways located at important locations, providing TCP/IP service throughout a large geographical region.

Figure 4. A repeater is mounted on a tower on Sumas Mountain. This rebroadcasts TCP/IP packets from the 56 Kbps radio network.

Multicast video conferencing was performed over a combined link using spread-spectrum radio technology, satellite communications, and multicast tunneling. Kasei tunneled multicast video data from the Telematics Research Laboratory, over the space-based link to the Federal Joint Alternate Site (JAS) in Cloverdale, BC, and then radiated that data via 900 Mhz spread-spectrum to a vehicle located in a field nearby. Thus a high-performance satellite connection was made available to an area around the receiving Earth station.

Figure 5. The Joint Alternate Site. The large dish located in the right of the image is the main spacelink to the Integrated Network.

Figure 6. Linking to the satellite network via spread spectrum radio. Notice camera on top of vehicle, connected to notepad. Videoconferencing exchanged from vehicle.

Figure 7. The directional antenna linking back to the digital wireless modem located on the roof of JAS.

JAS was provided with a multiple-node wireless communications gateway system that can allow the local intranet to link into external internets via various communications technology. Each point of wireless transmission is firewalled, for security. Two buildings at JAS were connected with spread-spectrum technology, with one site providing 56Kbps access, and the other providing satellite communications via two satellite links.

Figure 8. The routing system used to access the Integrated Network at JAS.

Virtual Institute Networking technology was tested over the VSAT (Anik E1 Ku-band) satellite communications link, the radio link, and a low-cost DirecPC satellite communications link.

TCP analysis software was used to graphically check the packet flow through the links. SatCom links we controlled over the network by connecting Kasei into the control facilities on the satellite modem.

Due to the advances in making the system TCP/IP compatible, and able to maintain connections to PDG OpenMath technology, the C3 load monitor will be used for link status monitoring.

1.4: Global access system tests

Javastations will be integrated into the system, and a DHCP service established that will allow booting of these network computers over each link type.

Deliverable: Verification of ability to access services from simple network computers than can connect to the integrated network without local user configuration, via each link type (conventional, wireless, space-based).

Results

JavaStation booting over conventional networking, even when not on the same subnet as the boot server, was established. A DHCP relaying system was placed on Kasei to ensure that boot requests could be fed to Nirgal for processing.

Experiments took place in which the JavaStations were booted using 56 Kbps packet radio networking. Initial booting performed well. However, the large user configuration file required by the present generation of JavaStations (1 megabyte) could not be transferred without significant stalling, aborting complete configuration after booting. However, PDG and Telematics are investigating the use of higher-performance radio networking systems that will make booting of JavaStations feasible over these connections.

Figure 9. Global access to the Integrated Network at JAS. Notepad accessing videoconferencing, JavaStation accessing Virtual Institute Network. Satellite modem and digital multiplexer at left.

The JavaStations were booted over a range of VSAT satellite communication links. These tests went perfectly and were performed in conjunction with the Communications Research Centre at the federal government's Team Canada booth at InterComm99 in Vancouver as well as during the Thunderbird IV emergency exercise. The JavaStations were then used as terminals to access the Virtual Institute Networking system, and the various disaster information systems on Kasei.

JavaStations were also booted over the 900 Mhz spread-spectrum links.

The resulting integrated network has thus demonstrated the ability to boot computers from high-performance servers located anywhere on the planet, and then provide access back to collaborative services, including computational services, on the remote network.

Figure 10. Integrated networking, from a truly remote location!

It is this technology that will allow users to access a radio network from a portable computer, relaying data back to a vehicle, and from there back to a satellite communication system. From there, services on high performance networks and high performance computers can be accessed from any location within range of the satellite.

Background

The PolyMath Development Group (CECM) and the Telematics Research Lab (CPROST) at Simon Fraser University are developing a high performance, next-generation, networking capacity within a high performance computing hardware context. Utilizing a mix of ATM networking, wireless networking, and various modes of satellite networking, it will facilitate the development of projects requiring both significant computing and networking resources. This project is aimed at the integration of the resources and development of status and control interfaces for the emergency information system VEMIS.

One of the PDG’s SUN Microsystems Enterprise 450 is currently being integrated into the Telematics Lab. It will act as the communications hub between the local network, broadband-capable Very Small Aperture Terminal satellite links, VEMIS wireless emergency network and broadband CA*Net2 ATM network. It will mediate between the various technologies, using them efficiently, and providing rapid access to high-performance computing resources.

Initially the facility will support deployment of crucial resources for disaster preparedness (information and collaboration), emergency alternative networking (in the event of loss of CA*Net2 connectivity), and centralized access to information and collaboration during a disaster. Remote sensing information will be integrated with use of HPC resources to predict damage, and warehouse important information. These resources will be tied together through status and configuration interfaces developed in Java using PolyMath technologies.

In addition, existing Web-based administration and information services will be upgraded.