Actian Acquires ParAccel - What's Ahead?
The combination will offer end-to-end big data solutions built from the ground up, says a company spokesperson. We look at what’s ahead post-acquisition.
Actian Corp.'s pending acquisition of ParAccel Inc. raises at least as many questions as it answers.
The first and most obvious is a question about execution on a pair of Very Big Deals. Earlier last month, for example, it completed its acquisition of data integration (DI) specialist Pervasive Software Inc. Weeks later, Actian announced its intent to acquire ParAccel. Like Pervasive, ParAccel is an established player with a mature product; integrating the assets or resources of two companies almost simultaneously would be challenging for any company.
The “integration” issue is a thorny one. VectorWise and ParAccel are two very different technologies. They're both nominally analytic database platforms, but ParAccel is a columnar massively parallel processing (MPP) platform; VectorWise optimizes for chipset features, core densities, and single system performance: it isn't an MPP platform.
This begs the question of what Actian plans in the way of “integration.” Something deep and thorough, reminiscent of what Microsoft Corp. undertook with its acquisition of the former DATAllegro Corp., or something looser and (in a virtual sense) logical, ala Teradata Corp.'s acquisition of the former Aster Data Systems Inc.?
Pervasive veteran Mike Hoskins, now CTO of the combined Actian Pervasive, says the initial plan is to emphasize a kind of tiered data warehousing (DW) stack.
“VectorWise can do very nicely up to the 10-20 TB range, but for the very large customers ... you need a much larger apparatus [i.e., ParAccel's MPP platform]. The acquisition is kind of a capstone in building out an end-to-end portfolio of ... very modern super-scaling technologies,” he says, adding: “[We're] probably [going to spend] the next 60-90 days appreciating and figuring out all of the latent power in this combination.”
Reading the Integration Tea Leaves
There are three potential integration scenarios. The first and most ambitious involves combining the distinctive strengths of both Actian's VectorWise platform and ParAccel's MPP technology in a single combined data warehouse (DW) platform.
Actian touts VectorWise's ability to scale linearly across very large SMP systems; the idea is that Vectorwise will run about twice as fast on 32 chip cores as it will on 16.
Unlike ParAccel and most other analytic database platforms, VectorWise isn't an MPP engine. True, it does exploit Intel's single-instruction, multiple-data (SIMD) technology to parallelize workloads on the same system, but VectorWise doesn't distribute workloads to run in parallel across an MPP cluster. That said, VectorWise's use of chip-based SIMD extensions and vector-processing capabilities was and to some extent still is cutting-edge. Its emphasis on large-scale SMP scalability was likewise prescient, inasmuch as it anticipated the shift to “Big Density” -- i.e., the trend among Advanced Micro Devices Inc. (AMD), IBM Corp., Intel Corp., and even the former Sun Microsystems Corp. to consolidate multiple chip cores into a single processor package.
An ambitious integration scenario would produce a combined scale-up and scale-out DW platform: a system that's optimized for SMP performance at the node level and which can scale out in large MPP configurations. This is something that Hoskins says Actian hasn't talked about all that much.
“The real end-game is to offer customers a smooth path with a single vendor and a single infrastructure vision where they can grow from their current needs to their future needs,” he indicates. At this point, Hoskins says, “We're not revealing anything. We're fast and furiously working to reach a more scalable solution. At the very least, we have loose coupling [between the two platforms] that works out of the box.”
Think of what Hoskins calls “loose coupling” as the most likely integration scenario. It emphasizes a stack-like architecture, with VectorWise positioned for certain kinds of workloads, ParAccel for others, and Pervasive, with its massively scalable DI technology, providing integration coupling between the two. This is what Hoskins and Actian claim they'll be able to deliver almost from the get-go -- once the acquisition officially closes. It's the scenario that many experts think is most likely, too.
“Actian did say they were planning to provide a scale-out solution for VectorWise and so maybe they will use ParAccel to do this. Also in addition to DI, Pervasive also brings Datarush Analyzer which is a very interesting product,” comments industry luminary Colin White, president of BI Research, who -- like many of his peers -- believes a combined Actian-Pervasive-ParAccel could be a formidable competitor. “They have all of the pieces to provide a fairly complete solution if they can integrate the components.”
The question is: integrate how -- deeply or loosely? A third, more pragmatic scenario involves an upfront effort emphasizing loose coupling, followed by a year-long project focusing on integrating the VectorWise and ParAccel platforms.
Hoskins repeatedly refused to speculate about this scenario.
“I don't think there's enough substance there for me to venture comment. Until you finish an acquisition, it's pretty much head-down running your business,” he said.
His circumspection isn't unwarranted. An integration project of this kind would likely be daunting: Microsoft's effort to assimilate DATAllegro's MPP technology took nearly three years, culminating in the 2011 release of its SQL Server Parallel Data Warehouse (PDW) product. Other vendors that have purchased MPP database companies -- e.g., EMC Corp., Hewlett-Packard Co., IBM Corp., and Teradata -- continue to resell them as more or less discrete platforms, tied via middleware or other technologies to their larger data management (DM) ecosystems. This is precisely what Hoskins says Actian has committed to do. Some in the industry don't expect it to go much beyond this.
“Reading all the viewpoints on the topic the general perspective is they won't try and integrate Vectorwise and ParAccel,” says a prominent data warehousing specialist with decades of industry experience. This industry expert, who spoke on condition of anonymity, believes that Actian will adopt a Lite Integration-like strategy: “I sense the strategy may be to sell VectorWise for small deployments and ParAccel for big ones.”
The Sun Never Sets on OLTP?
Veteran industry watcher Mark Madsen, a principal with information management consultancy Third Nature Inc., had another thought: one thing that isn't being talked about is Actian's OLTP database platform, which it inherits from parent company Ingres.
Madsen says Actian -- like DM giants IBM Corp. and Oracle Corp. -- has the makings of a tripartite OLTP, BI, and analytic database stack. According to Madsen, its VectorWise platform is ideal for heavy, scan-based SQL workloads, which typically run on a single large system. This addresses what he's described as the “basic BI” use case: e.g., batch reporting, ad hoc reporting, dashboards, and (to some extent) SQL-based analytics. The ParAccel technology gives Actian a best-of-breed MPP analytic database that addresses petabyte-like scale-out scenarios, Extreme SQL analytics, and analytic discovery.
What about the OLTP database? Ingres is by no means a dominant OLTP platform, but it is mature, functional, and reliable. Hoskins seemed to downplay this idea, however.
“That [i.e., the tripartite stack] does represent [a kind of acquisition] over-reach. We're very happy with our core businesses. They are tremendous producers of customer satisfaction. The sun has set on some technologies: I mean OLTP,” he explained.
“What has really changed is that data is being born digitally now in a way that it wasn't before. It's constantly flowing: the tap is on and it'll never be turned off again.”