InformationWeek
Defining The Business Value Of Technology
January 17, 2000
 Printer ready |
Technology offers data retrieval much faster than that of disk-based systems
By Charles Waltner
| Related links: |
|
|
 Send Us Your Feedback |
dmond Mesrobian, chief technology officer at CheckOut.com, has always known that personalizing entertainment information and Web pages for customers would improve the site's ability to sell its music, movies, and games. But Mesrobian also knew it would be difficult to build a dynamic data-cache system for real-time information feeds that would work with a relational database. And in the end, such a system might not provide the company with the performance it needed.So he took a different approach: a prebuilt in-memory database from TimesTen Performance Software Inc. In-memory databases can perform data retrieval many times faster than traditional disk-based databases, because data retrieval isn't hindered by the innately slower process of pulling information off disk drives.
In-memory databases address the growing need of Web merchants for rapid, high-volume data delivery to run everything from shopping carts and personalized Web pages to targeted banner ads. That means customers aren't turned off by Web pages that take too long to load, making in-memory database technology a viable option for Internet companies struggling to customize Web pages for a growing number of users.
TimesTen also offers the Beverly Hills, Calif., company a sophisticated way to handle data in RAM, as well as replication processes for copying and synchronizing data for improved reliability. "Building a cache in-house with those features isn't a trivial task," Mesrobian says. "People may not be knocking down the doors to get at this technology now, but in a year, as customization becomes more common on the Internet, they'll be needing it."
Nevertheless, the technology offers distinct advantages over its disk-based brethren. In-memory databases aim to provide far greater data-access rates by doing away with disk-related processing by managing all data off RAM. The approach avoids the speed limitations of pulling data off hard drives to redraw Web pages with new information.
Most Web sites trying to address that issue rely on data caching: leaving bits and chunks of the most relevant data in RAM for rapid retrieval by the Web server. But even the most sophisticated of these techniques are hard- pressed to match the performance advantages of in-memory stores. While applications and architectures influence performance rates, in-memory databases deliver Web information four to 40 times faster than traditional relational databases, say executives familiar with the technology.
Since all the information in an in-memory database is highly structured, as opposed to sitting in files or rudimentary tables in cache, the database is more adept with dynamic queries requiring a wide range of data at any one time, says Al Hilwa, a research director at Gartner Group.
That kind of performance is attractive to Egreetings Networks Inc., which was recently ranked by Media Metrics as having the eighth-highest hit volume of all Web sites. The company turned to TimesTen to handle the load. Don Chaney, VP of engineering for Egreetings, says TimesTen's in-memory database has completely cleared response bottlenecks.
Despite glowing reviews, in-memory databases have drawbacks. Since data is stored in RAM, it will be lost if machines go down. The technology and the vendors are new and small, so companies aren't guaranteed long-term support. And because the technology is immature, products offer few extra features for easy integration and management.
Companies that use in-memory databases are also committed to buying more RAM as they upgrade or expand the database. "No matter how cheap RAM gets, disk media will always be cheaper," Hilwa says. Also, in-memory databases are limited by RAM capacity. And implementing these tools isn't easy.
continued...page 2
Back to This Week's Issue
Send Us Your Feedback
Top of the Page
Featured Microsite
