Supercomputing upgrade to improve big data at Postal Service
- By Frank Konkel
- Oct 09, 2013
The U.S. Postal Service' supercomputing and big data efforts are about to get a substantial upgrade.
USPS has issued a five-year, $16.7 million contract to Maryland-based FedCentric Technologies to expand the agency's already formidable high-density supercomputing capabilities, now used primarily for fraud detection, to improve efficiency and competitiveness.
Currently, USPS operates a supercomputer with 16 terabytes of in-memory computing that is housed within Eagan IT and Accounting Service Center in Eagan, Minn., capable of scanning 528 million mail pieces per day against a database of 400 billion records in near real-time for purposes of visibility, reporting, sorting and fraud detection.
FedCentric Technologies will build USPS four more similar supercomputing systems at the Eagan facility, significantly improving the frequency, quality and quantity of data coming through USPS networks.
Individual mail pieces will be tracked up to 11 times as they route through USPS systems, and that data allows USPS to greatly improve efficiencies, according to Gerry Kolosvary, president of FedCentric Technologies.
"There is great intelligence information coursing through USPS systems," Kolosvary told FCW. "It allows them to be more competitive moving forward. We know what is in the mail carrier's bag the next day, and that leads to all kinds of useful decisions, like how many trucks to send out, how many people are needed for a given day. That's a lot to begin to look at."
Even small efficiency gains could prove enormously important for the cash-strapped USPS, which has an annual operating budget of close to $60 billion. After losing $16 billion last year, USPS expects to lose another $6 billion this year. USPS' big data applications could help it recoup some of those losses through efficiency gains.
"It improves efficiencies, and there are a lot of capabilities for USPS to save money and expenses and that sort of thing," Kolosvary said.
USPS' big data efforts surrounding fraud detection will be improved as well, performing as many as 2 billion scans per day. When a mail piece is scanned at a post office facility, its relevant data – carrier and routing information, weight and size – is sent via the Postal Routed Network to the supercomputing facility and compared against all existing records. Fraud attempts are reported to the U.S. Postal Inspection Service for further investigation.
"Today, the USPS uses our systems for visibility, reporting, sortation, fraud detection and deterrence," Kolosvary said. "The current systems established the ability to process streaming mail piece data in real time. Future efforts will leverage this capability for additional high speed analysis and detection."
Frank Konkel is a staff writer covering big data, mobile, open government and a range of science/technology issues. Connect with him on Twitter at @Frank_Konkel.