Amazon Redshift Leaves On-Premises Opening, Says ParAccel

Did ParAccel strike a Faustian bargain when it accepted $20 million in funding from Amazon in July?

That investment added a big chunk to ParAccel's total venture capital haul of about $85 million. But in the bargain, Amazon gained license rights to ParAccel's database, which is now the foundation of Amazon Web Services' (AWS) new Redshift data warehousing service.

ParAccel is in the business of selling database licenses for on-premises deployments, but with Redshift (now in limited preview and set for release early next year), Amazon says it will deliver an easy-to-consume service that will be one tenth the cost of conventional on-premises data warehouse deployments.

Announcing RedShift at last week's AWS re: Invent conference, Amazon senior VP Andy Jassy described how "expensive and complicated" data warehousing is for large companies and how "out of reach" it is for smaller firms. Delivered as a service, Redshift will cost as little as $1,000 per terabyte, per year. That will be "game changing" versus the estimated $19,000 to $25,000 per terabyte, per year that companies are used to shelling out for on-premises deployments, Jassy said.

[ Want more on Amazon's new big data option? Read Amazon Debuts Low-Cost, Big Data Warehousing. ]

You might think ParAccel is now thinking, "with investor 'friends' like these, who needs competitors?" But Chuck Berger, ParAccel's CEO, doesn't see it that way.

"Amazon is using our technology for a major service offering, and that gives us a much stronger aura of credibility to any customer who might be concerned about dealing with a small company," Berger told InformationWeek. That will help ParAccel win deals against its usual competitors: EMC Greenplum, HP Vertica, IBM, Oracle and Teradata.

As for the threat of Redshift cutting into on-premises sales, Berger launched into a litany of reasons why on-premises deployments will live on no matter how appealing Redshift might be. Topping his list are security, regulatory and performance concerns.

On security, Amazon insists it has all the corporate concerns covered with Redshift and its other cloud services. And Berger admits that plenty of large companies are comfortable working with Amazon. But in the financial services industry, where ParAccel counts Goldman Sachs and Royal Bank of Scotland as marquis customers, Berger said many companies "just aren't going to move their data or their applications into a public cloud."

On performance, Berger said on-premises offers advantages over cloud when it comes to getting data into and out of the database -- a key issue we raised after last week's Redshift announcement.

"It's obviously faster to load data in your own data center than it is to send it across even the highest performance connectivity solutions that Amazon has available," Berger said, adding that the best options from Amazon are expensive and "somewhat complex to set up."

The data-loading options in question are Amazon's Virtual Private Cloud and AWS Direct Connect. The former is a virtualized private network offering that Berger described as complex. The latter is a private connection between AWS and a corporate data center that costs $0.30 per hour for a 1 gigabyte-per-second port and $2.25 per hour for a 10 gigabyte-per-second port. To put that in perspective, the Redshift service starts at $0.85 per 2-terabyte node, per hour, so the best connection is more twice as expensive as the service -- though it could presumably be used in conjunction with a variety of Amazon services.

Another performance differentiator, said Berger, is the availability of high-speed connectors for Oracle and Teradata databases and Hadoop as a data source to ParAccel. These connectors have not been licensed by Amazon -- not yet, anyway, Berger said.

One final differentiator between Redshift in the cloud and ParAccel on premises is the level of customer support available. Where Amazon looks to standardize its services for broad consumption, ParAccel helps its customers exploit features designed to tune the database for each deployment for higher performance. "Those are support-intensive and professional-service-intensive features, and that's not Amazon's business model," Berger said.

InformationWeek has asked Amazon for follow-up interviews on Redshift. At press time the company had yet to respond. [[Author's update: Amazon sent answers to our questions -- posted below in the "comments" area -- four hours after this article was published.]] We told them we wanted to hear more about data-loading options and speeds. We also raised questions about just how and where companies will run their business intelligence (BI) software.

Running BI in the cloud will add to total cost. Jaspersoft and Microstrategy are the first Redshift partners in the cloud. Even if you're using Jaspersoft's Community edition, which is free from a software licensing perspective, there are Amazon charges (ranging from $0.16 to $1.80, depending on compute power) for provisioning the software-as-a-service. If you stick with running BI in your data center, you have to remember that bringing data back on premises for analysis will introduce additional delays over and above the time required to load data into the cloud.

Perhaps Amazon is working out the details, but we're still hoping to learn more about Redshift data-integration and BI workflows before the service goes live early next year. Until then, ParAccel says the ultimate in security, control and performance on its database can only be achieved on premises.