By Jarrett Cohen
NASA-internal wide-area networking services for the High-End Computing (HEC) Program are transitioning to the NASA Integrated Services Network (NISN) on Wednesday, May 27, 2009 at 1 PM Eastern Time/10 AM Pacific Time. HEC's new NISN services include a direct 10 gigabit-per-second network pathway between the NASA Advanced Supercomputing (NAS) facility at Ames Research Center (ARC) and the NASA Center for Computational Sciences (NCCS) at Goddard Space Flight Center (GSFC).
Once the cutover to NISN takes place, service currently provided over the NASA Research and Engineering Network (NREN) to NAS and NCCS will automatically be routed over NISN links. This switchover will be transparent to users. The cutover does not affect HEC traffic to and from the Jet Propulsion Laboratory (JPL).
"We formed a strategic partnership with NISN to consolidate our operational wide-area networks with those across the agency," said Tsengdar Lee, HEC Program Manager at NASA Headquarters. "This move is in keeping with NASA's new coordinated approach to information technology, which minimizes duplication and accelerates the infusion of new technology into the operational network."
NISN handles wide-area network traffic between all NASA centers and connects to other federal networks, Internet2 (primarily universities), and the broader Internet. NISN and its predecessors have served as the agency's primary networking infrastructure for 20 years.
The NASA Integrated Services Network (NISN) added its first 10 gigabit-per-second (Gbps) link for traffic between the NASA Advanced Supercomputing (NAS) Facility at Ames Research Center (ARC) and the NASA Center for Computational Sciences (NCCS) at Goddard Space Flight Center (GSFC). Figure courtesy of NISN. (Click on the image for a larger version including acronym definitions.)
At 2.5 gigabits per second, the current NISN backbone is too slow to meet HEC's unique requirement for cross-center data backups, which involves sending 2 terabytes each way between centers on a daily basis. This capability ensures that data is available at a second physical location in the event of a natural disaster or other unforeseen event. Related projects in the planning stages include user file transfer and a wide-area file system.
"We had the first requirement for 10 gigabits per second in the NASA family," said Pat Gary, Networks Manager in GSFC's Computational and Information Sciences and Technology Office (CISTO)—the NCCS parent organization. After the decision was made to consolidate networking services, HEC and NISN collaborated on building a custom service for carrying HEC production traffic between NAS and NCCS. "The NISN link is tailored to meet our needs," Gary said.
Langley Research Center (LaRC) also partnered with HEC and contributed to the transition. This partnership resulted in NISN quadrupling the speed of LaRC's links up to a total capacity of 5 gigabits per second. In addition to other uses, the upgraded links can better handle the network traffic of LaRC scientists and engineers carrying out massive simulations at NAS, which include operational support for Space Shuttle flights.
To ensure delivery of high-performance and high-availability networking services after the transition, CISTO, NAS, and NISN network engineers established an ongoing collaborative testing and troubleshooting strategy. The process began with 50 individual acceptance tests, which investigated areas including raw network performance (i.e., sustained speed), data packet error or loss rate, and jitter (i.e., the variation in arrival times between packets). Engineers tested packets across the pathway between NAS and NCCS alone, across that pathway with the addition of LaRC-to-NAS traffic, and across the NISN failover pathways.
"We stressed the pathway between NAS and NCCS really hard, pushing it harder than any user will," Gary said. "We banged away on this thing, and it really came through." Network performance was as high as 9.3 gigabits per second (out of a theoretical maximum of 9.6 gigabits per second) without losing any packets. Several acceptance tests required continuous transmission for 48 hours, and again not a single packet was dropped.
Continuous performance monitoring and various user tests have been taking place since last summer. Every 15 minutes, workstations at NAS and NCCS perform and measure the speeds of memory-to-memory transfers between the two centers. For specific user applications, NAS and NCCS mass storage experts tested the cross-center backup capability over the new pathway. Additional tests between LaRC and NAS and the Jet Propulsion Laboratory and NAS included bidirectional memory-to-memory and disk-to-disk copying using several file transfer protocols. Across these different configurations and applications, transfer rates over NISN have been comparable to those on existing providers.
As before the transition, users who have connection or transmission problems should contact NAS or NCCS User Services. Kevin Jones, NREN Project Manager at NAS, also stressed that users need to be clear about their requirements. "If there is an impact on getting their work done, they need to escalate that to their appropriate people," Jones said. He cited management at NASA centers, mission directorates, and CIO organizations, who would then need to communicate with HEC. In turn, NISN receives all of its requirements from HEC. A Service Level Agreement between HEC and NISN establishes the official requirements.
The recent upgrades to the NISN network in support of HEC requirements are early deployments of the larger Wide-Area eXpansion (WANX) project, conceived to expand the NISN backbone to 10 gigabits-per-second, with regional feeds also expanding in line with NASA user requirements.
NISN's Wide-Area Network Expansion (WANX), scheduled for 2009, will upgrade the NASA backbone network to 10 gigabits per second (Gbps). Figure courtesy of NISN. (Click on the image for a larger version including acronym definitions.)
For the pathway between NAS and NCCS, "there were some new configurations that we had to undertake," said NISN's Carol Bryant, Project Manager in the Office of the CIO at Marshall Space Flight Center. "All the technical details of bringing this first 10 gigabit-per-second link into production operations will definitely help us with the next phase of the WANX project."
For more information about NISN, visit:
For NAS and NCCS User Services contact information, visit:
For slides showing acceptance test specifications and results, contact:
Pat Gary, email@example.com