Feds tiptoe toward in-memory computing
- By Frank Konkel
- Sep 24, 2013
It is not yet as mainstream as other technologies such as cloud computing and big data, but federal agencies are beginning to make use of in-memory computing.
While many federal agencies have adopted cloud computing for e-mail and other basic services, and a few have started legitimate big data initiatives, the high cost and uncertainty of returns of IMC platforms previously kept all but the most cash-flush organizations or agencies from exploring the technology.
But those that have used it are happy with the results.
One large U.S. law enforcement agency recently adopted Enterprise Ehcache, an in-memory data management platform offered by Software AG – the parent company of developer Terracotta – that provides the agency with a 500-gigabyte in-memory data store, complete with an analytical engine and SQL-like search capabilities.
The agency, which Software AG Government Solutions is not allowed to name, turned to IMC to deal with a 10-fold growth in the volume of data it has ingested daily since 2011; it now receives more than 100 gigabytes per day, or the equivalent of 1 million new records. The agency's 500-gigabte in-memory data store allows new information to be accessed and queried almost immediately by agents in the field or analysts in the office.
"What we're opening up for this customer is the ability for them to have a tremendous amount of information at their disposal, in-memory, and that means real-time," said Bill Lochten, vice president of Software AG Government Solutions.
"We're able to improve performance in that we can provide information back to agents in the field in the moment," Lochten said. "You can think of an agent literally having to verify the status of an individual or pieces of information about a potential criminal matter – we provide a 360-degree view of the situation in microseconds rather than minutes."
The system can handle requests to its central repository from thousands of users at once. If a request for information isn't available in the in-memory system, the IMC database pulls the data from main storage for further processing or access, said Fabien Sanglier, solution architect for Software AG Government Solutions.
If multiple users make similar requests, the system recognizes it and saves hot data sets in-memory, saving time for future searches. This can be important given the sometimes disparate nature of criminal databases, Sanglier said. If a search query on the suspect takes minutes or longer instead of milliseconds, that's potentially a very big problem for law enforcement.
"Having the ability to see the complete picture while in the moment, that's what is most important to law enforcement," Sanglier said.
Sanglier said the agency's IMC is scalable to the entire memory of the physical servers or virtual machines on which it runs. And since Enterprise Ehcache runs on existing hardware, Sanglier said, the solution itself is scalable as the user continues to consume more data.
While the government is still in the early stages of adopting IMC and its various hybrid models, the velocity and volume of data feds produce, analyze and store continues to grow.
According to Gartner, its future in the federal market is linked to how feds decide to use the data they have. Agencies within the intelligence community (IC) already use IMC for a variety of computing jobs that require low-latency results. IMC could seem logical to civilian agencies that work regularly with increasingly large amounts of data, such as the enormous forecast models and climate data derived by the National Oceanic and Atmospheric Administration.
"IMC opens unprecedented and partially unexplored opportunities for business innovation (for example, via real-time analysis of big data in motion) and cost reduction (for example, through database or mainframe off-loading)," according to Gartner's "Hype Cycle for In-Memory Computing Technology 2013," published in July.
"In-memory computing technologies are moving fast toward becoming mainstream, driven by rapid hardware evolution, maturing software cloud deployment models, the proliferation of innovative use cases and software packages embedding these technologies," the study said.