In-Depth
Sybase Touts a Faster, More Scalable IQ
With IQ 15, Sybase takes aim at columnar competitors ParAccel and Vertica -- as well as the teeming MPP appliance market.
Sybase Inc. last month unveiled a new version of its venerable Sybase IQ analytic database. With advanced query algorithms and improved query parallelism -- to say nothing of grid capabilities, a beefed-up management console, and support for FIPS, Kerberos, and IPV6 -- Sybase IQ 15 amounts to a data warehouse reloaded, Sybase officials say.
It's a timely release, too. Sybase IQ is no longer the only columnar database available. Competitors ParAccel Inc. and Vertica Inc. both market what they claim are newer, fresher spins on the columnar data warehouse. To a degree, both players like to position Sybase IQ -- politely, respectfully -- as old hat.
Sybase officials, not surprisingly, reject these claims, touting IQ's longevity as a strength -- particularly in today's business climate.
"A big part of how we distinguish ourselves from other columnar players is just the fact that we're the established player. We have credibility that they can't offer [with more than] 3,000 implementations world-wide," says Joydeep Das, senior product manager with Sybase. "It's one thing to get a technology out in the market; it's another thing to have companies bet their businesses on it. With IQ, our customer service satisfaction is over 70 percent. For a company of our size, that's unusual."
Das, like his counterparts at both ParAccel and Vertica, talks up the advantages of (in Sybase's case) IQ's columnar architecture for analytic workloads. In the current climate, he argues, businesses will survive or thrive based not just on the strength of their analytics, but on how they use their analytic insights to drive decision-making.
"Our positioning is the sort of unique place where we are with this product in the analytics space: we offer unlimited scalability in multiple dimensions. That's really what we offer with IQ. That's part of what differentiates us, particularly when you couple that scalability with the flexibility that we offer," he says.
On the other hand, Das concedes, IQ's columnar-cum-lately competitors claim to offer similar performance and scalability. How does Sybase propose to differentiate itself? "The difference [with IQ] is our ability to provide answers so businesses can meet their customers' requirements and respond to rapidly changing business conditions," he comments.
At any rate, Das maintains, Sybase IQ performs much faster than traditional DBMS platforms from IBM Corp., Microsoft Corp., and Oracle Corp. It enjoys a similar performance advantage over Adaptive Server Enterprise (ASE), Sybase's venerable RDBMS. It's a now-familiar refrain: ASE and other conventional DBMSes are optimized for OLTP workloads, Das says. For years, customers have gotten by -- basically, treading water -- using these platforms for data warehousing workloads. Increasingly, he argues, the size and scope of their analytic requirements are flat-out outstripping the capabilities of these DBMSes, however.
"They're finally hitting the wall. If you throw enough CPUs and DBAs and indexes and money at one of these OLTP environments, you can squeeze some additional performance out of them. But if you're stuck in the mindset of 'we are a fill-in-the-blank shop and we can only use that database,' you're doing to be left behind," he says. Das decries the use of fixes that he says are designed to paper over -- not to properly address -- query performance and scalability issues in traditional IBM, Microsoft, Oracle, and even Sybase ASE DW environments. "Companies have invested heavily in standards of technology and databases, so it's not easy for them to yank this stuff out, so they use things like materialized views … [and] cubes as sort of Band-Aids on top of these transactional systems to get over their limitations."
It's a familiar argument, but Sybase claims to solve it in an unfamiliar way: in place of the brute MPP horsepower of the DW appliance vendors -- and the columnar MPP proposition of both ParAccel and Vertica -- it trumpets the strengths of IQ's supplementary stack, which includes best-of-breed data replication (Sybase Replication), grid computing (based on the former Avaki technology), and data modeling capabilities (Sybase PowerDesigner), along with a credible ETL toolset (which Sybase acquired three years ago from the former Solonde AG). These products are still sold separately; that could change in upcoming revisions of Sybase IQ -- although neither Das nor his colleague, Phil Bowermaster, worldwide product marketing manager for Sybase IQ, were willing to go into specifics.
"When we made this shift to addressing specific markets with Sybase IQ, that turned some lights on in terms of how to position the data integration suite. Having the whole thing [all of the data integration capabilities] in a box sounds great, but when you go talk to individual customers, they have a particular set of needs and they're looking for particular capabilities to meet a particular business challenge," Bowermaster comments. He demurs on the question of whether Sybase plans to bundle its creditable DI stack with IQ.
In the near-term, Sybase is content to position its DI tools as complements to IQ. That alone is a value proposition that its MPP or MPP columnar competitors can't match, Bowermaster says.
As for the IQ engine itself, there are plenty of new features in version 15 to keep existing customers happy -- and, Bowermaster contends, to entice new customers, too. "IQ [version] 15 is completely multicore aware, and we're parallelizing everything -- starting from scan after data set to relational operators. We're also doing all of that in a vertically parallel fashion," he points out. He is proud of IQ 15's revamped management console -- it now boasts integration with both Sybase WorkSpace and Sybase PowerDesigner, while also supporting one-click grid configuration and deployment capabilities -- and boosted security services.
Couple that with IQ's improved query performance -- Bowermaster points to sub-query performance improvements, enhanced in-memory query processing, and improved tokenization -- and Sybase believes it has significantly chipped away at the value proposition of the MPP appliances.
"In many ways, it takes the momentum away from MPP in certain respects. With our technology, we can not just attack each box, but we can string these boxes together in a grid, so we can provide a narrow band of performance irrespective of how many users come on to the system. If you need to add just storage, you can just add additional storage. If all you need is compute [capacity], you can just add that," he argues.
"We're not throwing million-dollar boxes at the problem. The thing with MPP architectures is that they're great, they really work, but if you want to scale out, you have to invest in both storage and compute devices at the same time."