Data Management

STARTUP TO WATCH

ParAccel’s Bid to Build a Better Data Cruncher

Data mining is becoming a crowded field filled with software providers using similar strategies. Their basic goal is always the same. The analytics platforms are designed to slice and dice data to make sales trends and buying opportunities more evident.

The firms that can deliver this product more accurately and more rapidly grow their reputations and entice new customers from lesser-producing competitors. Even when relative newcomers such as ParAccel introduce innovations, what they provide to their customers is essentially a better, faster way to dissect digital data.

For example, ParAccel announced in early December its next-generationScalable Analytic Appliance II. This latest hardware appliance is intended to boostperformance and efficiency by integrating flash-based server arrays.

ParAccel’s founders are Barry Zane, cofounder of database heavy Netezza; CEO and President David J. Ehrlich, former senior vice president and marketing and chief strategy officer at NetIQ; and Bruce Scott, former vice president of Engineering at SenSage. Earlier, Scott was one of four cofounders of Oracle, serving as the principle engineer and architect of the Oracle database’s first three versions.

“ParAccel has been around since 2006 in a very crowded and fast-growing market that quickly deploys business intelligence. It has data warehousing appliances with pre-configured software to deliver its Web analytics specialty via a variety of platforms,” James Kobielus, senior analyst at Forrester, told TechNewsWorld.

Hanging On

Perhaps the trait that most distinguishes ParAccel from some of the largest data mining vendors in the market — think Oracle and IBM — is its software, which uses a columnar data base. This is a different way of structuring the database for faster response. The information is structured by columns rather than tabs.

With just about a dozen customers — including OfficeMax, Merkle, Fidelity National Info Systems, Nielsen AIG and Price Chopper — ParAccel is competing against a number of companies with a much larger customer bases.

However, the columnar database concept may help draw more in. It can be deployed as a stand-alone product or in front of an existing database to speed up the results. This approach saves on storage costs, and its ability to quickly load data enables it to handle terabytes per hour.

“It’s fast due to the strong compression and highly parallelprocessing platform optimized for EMC, Sun and HP platforms,” Kobielussaid.

A Better Mousetrap?

Many companies today have more data than they can process. They’re drowning in data they can’t mine without costly and time-consuming manipulation. Extracting intelligence from raw data requires significant man-hours to write programs for complex queries and then days more to run them.

Seeing these limitations first-hand drove ParAccel’s founders to develop a different approach to analytic database management, one they believe enables companies to ask virtually any question of their data on the fly and get an answer in minutes. Companies really will be able to respond in real time to changing market trends and consumer patterns.

“The idea to do commodities-type analysis on hardware began when I was still at Netezza. I was against that idea with that company. Its market strategy was working well. If something is succeeding, you don’t mess with it,” Barry Zane, CTO of ParAccel, told TechNewsWorld.

Early Moves

However, the idea still held Zane’s interest, so he agreed to join the other founders and pursue a company to develop that product. He moved to San Diego to set up a base of operations and pursue initial funding.

With enough cash in hand from a capital venture source, Zane set out to put a team of 18 engineers together. That task proved daunting.

“It was a task to find young, energetic engineers who weren’t polluted in orthodox thinking. It takes a lot of engineering talent to solve problems involving large systems,” said Zane.

Salvaging Success

The defining moment for ParAccel’s proof of concept came some four months into the new company’s product development efforts, noted Zane. That was back in 2005, when Zane was still showing the company’s potential as a proof point to get early venture funding.

The development team was running its proprietary software on 48 servers. The heat factor set in. Zane remembers scampering around to move the server farm into a larger room. He had to run to a local store and carry out a number of large floor fans to cool the overheating servers.

Such incidents are typical of startups, Zane quipped. He emphasized that his engineers solved that problem and no longer have that issue in any of its appliances. ParAccel products now run on much larger server systems.

Quality Control

Another incident that Zane sees as defining moment was his new company’s reliability for high performance involved a POC (Proof of Concept) bid he submitted. The engineering took 10 days, and led to a contract offer — but he declined it.

“We had the best performance but knew we couldn’t meet [the potential client’s] long-term demands,” he explained.

Changing Spaces

Zane and his fellow cofounders saw a change in the marketplace and aimed their product to meet that niche. The nature of what people do with their data is evolving, he noted.

“Traditionally, companies make very simple analytical questions. For example, a store chain will want to know what products sold well at a particular time of the year. Clearly, a more interesting tend is the correlation of information,” said Zane.

For instance, modern day data mining activities push the envelop on the concept of market basket analysis.

Did you ever notice a beer display teamed up with baby diapers in a store? How do the store’s marketers know which products to team up? That’s what data analysis is all about today.

Figuring Futures

From Zane’s perspective, ParAccel has a clear lead in understanding how to build these columnar database systems.

“We’ve never been beaten on performance. We don’t leave road kill behind,” said Zane. “I don’t see that changing.”

Name recognition is not a problem; he has deliberately kept ParAccel fairly low-key in the media. “It took solid engineering to get here,” he said.

That approach will allow the company to make the most of success in the $6 billion information market place, he said.

Hardware Endeavor

In what could be the next big step forward in reaching future goals,ParAccel announced earlier this month its adoption of new, fast flashstorage technology. The ParAccel Scalable Analytic Appliance II (SAAII), the company’s latest hardware appliance, incorporates the speedand efficiency of flash storage technology with breakthroughs inperformance optimization and enterprise manageability of the newParAccel Database (PADB) 2.0.

SAA II is designed for companies seeking massive scale and performancefrom their data warehouses while leveraging the enterprise-classcapabilities of SAN environments. The appliance is designed todramatically improve the performance of SAN-attached data warehouses.

SAA II is powered by EMC and ParAccel. It dynamically balances I/O andstorage across direct-attached compute nodes and the SAN to achievegreater I/O throughput and CPU utilization.

Leave a Comment

Please sign in to post or reply to a comment. New users create a free account.

More by Jack M. Germain
More in Data Management

Technewsworld Channels