A new big data solution promises to do more than help businesses analyze massive amounts information. It's actually "breaking the laws of physics," said Eldad Farkash, co-founder and chief technology officer at SiSense.
SiSense, a three-year-old developer of big data analytics software, announced patent-pending technology at the O'Reilly Strata + Hadoop World 2013 conference in New York City this morning.
SiSense claims it has mastered "crowd accelerated analytics": an innovation to SiSense Prism that provides business users with faster query results as loads — and the numbers of users — increase. "We call it power querying," said SiSense CEO Amit Bendov.
More is Better
While other solutions get "clunky and costlier" as load increases, SiSense Prism gets faster and "overcomes the increased expenses associated with slow-downs," Bendov added.
Until now, Farkash explained in an interview yesterday with CMSWire.com, the conversation has been around scaling data. The larger question, however, involves scaling massive amounts of users and queries. Traditional business intelligence solutions provided users limited answers to limited questions. But crowd accelerated analytics lets you analyze billions of rows of data in a flash. he said.
What's it mean? According to a company press release:
Crowd Accelerated Analytics is the only big data analytics technology to speed up as the number of users increases (more users = faster results). Traditional BI tools rely on caching technology that only works well when repeat queries are identical. But business users do not often ask identical questions of their data. While other solutions slow down or choke under the strain of increasing users with unique queries, SiSense Prism increases efficiency and speed by learning from results of similar but not identical queries."
It's another leap forward for Prism, which SiSense revealed at last year's Strata conference in New York City.
Prism is software capable of handling large analytic workloads on a single machine — a big-data analysis tool that the company maintans requires neither the skill set nor the budget of traditional analytic setups. (The cost: around $2,500 for a single annual license, including support and upgrades.)
The company demonstrated the capabilities of the ready-out-of-the-box software by processing a terabyte of information on a Dell Vostro 3560 decked out with 8 gigabytes of RAM. The laptop used could be purchased for less than $750 and the software was designed to run on most stock models.
Leveling the Playing Field
Bendov boasted that it leveled the playing field because it circumvents the traditional big-data requirements. It also makes sophisticated technology available to even the most technologically inept and financially constrained businesses. But a funny thing happened along the way. A lot of big companies — think eBay, Target, Merck — also saw value in the platform.
Prism features automatic extraction from all major databases (Oracle, Microsoft SQL Server, MyServer, and so on). It can store and analyze billions of rows of data, presents analyzed data in neat visual displays and completely cuts out the need for scripts of any sort.
In the past year, interest in SiSense — and its solutions that maximize the disk, memory and CPU on even small computers in the name of low latency — has soared. The company, with headquarters in both the US and Israel, raised $10 million in a second round of funding and reported subscription revenue climbed 520 percent year-over-year.
10 Terabytes in 10 Seconds
Much of the interest stems from the ongoing evolution of Prism. At a Strata Conference in Santa Clara, Calif. this past February, SiSense demonstrated it could analyze 10 terabytes of data in 10 seconds using a single run of the mill $10,000 Dell server node. As a post in PandoDaily explained:
SiSense’s Prism platform and its Elasticube analytical database maximize computational power by using a combination of in-memory, in-chip, and disk-based processing. Combining this hardware utilization with columnar databases and ETL (extract, transform, and load) tools allows the company to process data 50 to 100 times faster than its competition, legacy BI providers like IBM and Oracle. Moreover, it achieves this performance using commodity, “off-the-shelf” hardware. Individually, none of these techniques are unique to SiSense, but in combination, they put the company’s solution in a class of its own."
Now SiSense claims the technology is even better, with the ability to net faster and smarter query results as users increase. In addition to its ability to scale terabytes of data, Prism has the capacity to scale to thousands of users on a single box.
Farkash said costly bottlenecks are avoided and users benefit from each other’s queries even when queries are not identical. "We're slicing up the data in such a way that the query will stay in the CPU cache for reuse by additional queries," he said. (The CPU cache is the memory on chip that is used for temporarily holding data being processed by the processor.)
As Farkash explained, SiSense solutions depend primarily on disk for storage. They only move data to RAM as necessary, and then Prism uses vectorization and optimized instructions to handle as much work as possible in parallel on the processor, he said.
SiSense will be demonstrating its crowd accelerated analytics at its booth at the Strata conference this week.
But there was more from SiSense today than news about its technology. The company also announced it was swapping its offices in Redwood Shores, Calif. for a US headquarters on Wall Street in New York City. Could this fast growing big data company be positioning itself for a US initial public offering?
It's certainly within the realm of possibility: SiSense investors include Battery Ventures, which has supported multiple breakthrough companies, including seven who launched IPOs in the past fourteen months.