Whatever you call InterContinental Hotels Group (IHG)'s new data warehouse appliance, don't call it "entry-level" -- at least not when Alex Grigorian, vice president of enterprise technology
According to Gartner Inc., nearly all enterprises have a data warehouse these days, thanks in part to the introduction of lower priced, entry-level data warehouse appliances from vendors like IBM, Oracle and Teradata.
Teradata's new 2550 data warehouse, which IHG is in the process of implementing, is one such appliance, Gartner says. Not so, says Grigorian.
"With all due respect to Gartner, I'm hesitant to use the word 'entry-level' [to describe the Teradata 2550 appliance]," he said in an interview last week.
To Grigorian, who last November decided to replace IHG's five-year-old Oracle data warehouse with the Teradata 2550, the decision to go with an appliance versus a traditional enterprise data warehouse was a philosophical one, not a question of choosing a simple, beginner strategy over a more complex, advanced one.
"I am a change agent," Grigorian said. "This is a case where thinking out of the box will hopefully differentiate us from our competition. This will be a business intelligence tool that will be almost bar none in the industry."
In fact, IHG's data management needs can hardly be called entry-level. The U.K.-based hotel chain collects guest and reservations data from each of its 4,100+ properties and loads it into the Oracle data warehouse for analysis. Grigorian estimates that the total surpasses 10 terabytes of data.
In the past, as the amount of data it collected increased, IHG simply added more software and hardware to the Oracle data warehouse as needed. But soon after taking over responsibility for enterprise data management for IHG in 2008, Grigorian decided he wanted a more scalable solution.
"What I wanted to do was stand back and take a holistic view of the problem," he said. "The technology that got us to where we were [the Oracle data warehouse] was fantastic, but is it the technology that will get us where we need to be in the next 12, 26, 60 months?" The answer, it turned out, was no.
After evaluating both Netezza and Teradata data warehouses, and consulting colleagues at Bank of America who use technology from both vendors, Grigorian decided to go with Miamisburg, Ohio-based Teradata. But IHG didn't choose to deploy Teradata's enterprise data warehouse, the 5550, as most large enterprise customers do, but instead chose the vendor's smaller 2550 appliance.
The Teradata data warehouse appliance is more than capable of handling IHG's 10+ terabytes of customer data, Grigorian said, and it is scalable, thanks in part to its massive parallel processing (MPP) capabilities, so it can adapt to increasing data volumes. Grigorian said IHG plans to add financial data, and possibly employee data, to the appliance in the coming year for analysis.
But unlike a traditional data warehouse, the Teradata appliance will not need continuous performance tuning and optimization once up and running, he said. Once IHG completes loading its data into the appliance, Grigorian expects to enjoy its more robust analytic power "right out of the box."
One drawback of preconfigured data warehouse appliances, some argue, is that they are difficult to customize. What you gain in ease of use, deployment and administration, you lose in flexibility to make adjustments to changing data needs.
But Grigorian doesn't see it that way. "The word customization is an interesting one in this context," he said. "Do we call those customizations or do we call those necessary evils?"
"The appliance does not require [performance tuning and optimization]," he said. "An appliance brings that 'black box' environment where you don't have to stick your nose into the bowels of the system.
"But you have enough access and enough heft in the system to do whatever you need to," Grigorian said. "A lot of those tuning and optimization things that are always more traditional in RDBMS [relational database management system] environments, there will be no more with the appliance."
Another potential downside of data warehouse appliances is that, generally speaking, they cannot handle mixed workloads as efficiently as more traditional data warehouses. In the case of the Teradata 2550 appliance, it is also not optimized for real-time data loading, according to Dan Graham, product marketing manager at Teradata.
"The [Teradata] 5550 Active Data Warehouse would outperform the 2550 [appliance] when running mixed workloads," Graham said. "So we don't advise running near-real-time data loading concurrently with complex reporting on the 2550 appliance. It works, but not nearly as well. [The 2550 is] optimized primarily for reporting and analytics -- and priced for that workload too."
On that point -- price -- Grigorian does agree with Graham and Gartner. Though he declined to give exact figures, he said that "price was definitely a major decision maker" for IHG in selecting Teradata over competing vendors, and an appliance over a traditional data warehouse.
Going forward, if IHG's data volumes do exceed the 2550 appliance's capacity, Teradata assured Grigorian that an upgrade to its enterprise data warehouse, the 5550, would be a relatively simple one. Not that he expects that to happen.
"We are going to start sucking more data into our data warehouse [appliance] through various channels that will enable us to produce more intelligent reporting," Grigorian said. "I feel that an appliance will serve our growing needs. However, time will show."