Aster Appliance Elevates MapReduce Chatter, ADBMS Visibility

Since my last post about Aster, the analytic DBMS (ADBMS) vendor has added another arrow to its quiver. Its new MapReduce Data Warehouse Appliance Express Edition starts at $50,000, and includes Aster nCluster on Dell hardware and a copy of MicroStrategy BI software for up to 1 Tb of user data, which Aster clearly sees as a sweet spot. (MicroStrategy has been doing a lot of seeding with the ADBMSs lately; it also has  an introductory bundling deal with Sybase IQ.)  Delivering a ‘compute rich’ appliance on commodity hardware, with reduced operating costs, certainly hits all the right notes. But is 1 Tb  the sweet spot for MapReduce? I think not – although it makes a great starting point, and that may be Aster’s real opportunity – give ’em a taste of what SQL plus MapReduce can do, and watch them demand more and more. And sell it to them. Dell and MicroStrategy should love this strategy – if it works.

Even for those smaller data warehouses, speeds will be clearly improved. Lower first costs, ease of setup and administration – lowering both capital and operating expense – will lower the barriers to entry. $50K is a far cry from the half-million it can cost to get into other appliances from the “big boys.”  Once the value is proved, stepping into Aster’s Enterprise Edition, which it claims will scale to the petabyte range, may be easier to take. There are certainly some questions:

  • What’s the difference between a “data mart” and the “smaller data warehouse?”Aster quotes Gartner’s Donald Feinberg about the latter in its press release.  Perhaps Aster is choosing to ignore the data mart moniker – although it’s also possible that they are saying the improved generalized analytics of SQL plus MapReduce make it less necessary to restrict subjects and dimensions and follow specific architectural models the way many data marts typically do. If so, that will prove to be an interesting debate.
  • Are fault-tolerance and availability now “table stakes” for appliances? Aster is claiming “99.99% uptime, with reduced troubleshooting costs.” ParAccel has touted its relationship with EMC for enterprise-class “-abilities.” Other ADBMS vendors will need to keep up their features – and their rhetoric – here.
  • Is “SQL plus MapReduce” better enough to be a difference maker?Aster says that its “integrated SQL/MapReduce framework for analytics and BI increases query performance by 9x or more when compared with other SQL-only data warehouse appliances in the market.” Aster has not offered specific comparisons that show how the leverage of the two results in generalized improvements for particular use cases. It may be credible, but it certainly has not been shown yet – specific comparisons will hopefully be forthcoming.

Still, kudos to Aster for upping the heat, if not the light, in the emerging ADBMS wars. Aster opened a big door when it made MapReduce available to .NET, and no doubt some intriguing work will emerge from that community. Aster has a nice war chest to work with from its recent $17M Q1 financing round, and is putting it to work. So far the rhetoric has been aimed at Oracle, DB2, Teradata and Netezza. Easy targets. What about  Greenplum, Infobright, Kickfire, ParAccel, Sybase IQ, Vertica…? It’s going to be fun watching the smackdown ahead.

Published by Merv Adrian

Independent information technology market analyst and consultant, 40 years of industry experience, covering software in and around the data management space.

14 thoughts on “Aster Appliance Elevates MapReduce Chatter, ADBMS Visibility

  1. “…Aster’s Enterprise Edition, which it claims will scale to the petabyte range…”

    I generally recommend against measuring “scale” using how much data can be stored inside the DBMS (or ADBMS as you like to call it :D). See What I like about Aster is that they seem to agree, and a lot of their messaging around this appliance is how much processing power exists per TB of data. If you’re going to be pushing complex analytics into the ADBMS (e.g., with MapReduce), you need processing power.

  2. Thanks as always, and I highly recommend readers follow the link to Daniel’s post (and another one inside it!) I talk about “things vendors say” here a lot, and user data volume tends to be one of them, so I feel compelled to reproduce it. But your point is well taken, of course. Some benchmarks are designed to try to capture some of this subtlety as you know; when I don’t have that info, I’ll mention assertions or claims and try to make sure that I label them appropriately.

  3. Merv, excellent post as usual. As you point out, I’m not sure how MR fits in with data volumes at that level. I’m not sure either where Aster is anymore. Are they SaaS, are they appliance, are they software? Are they all over the place with mixed messaging? That’s my feeling at the moment.

    On the costing issue, I’m not impressed. XSPRADA’s software-only offering is at $25,000 up to 1TB as well, with free 30-day trial from – (and I challenge anyone to get anything including Aster up and running faster) — For a 5MB Windows service, I’d say that’s a pretty compelling offer. Oh and, we’re not in the cloud and we’re not an appliance 🙂

    Sorry for the marketing push but when I see stuff like this, it makes me go HUH?!@ and I feel compelled to evangelize once more 🙂

    1. Not to be a smartass, but your home page says:

      Learn more about
      our technology >Sign up to receive a limited-time free trial the moment the XSPRADA Self-Tuning Data Mart is released.

      So you need to update that. But YES – I’d love to set up a briefing.

  4. You are totally correct, my bad 🙂 We actually released Cinco de Mayo so, clearly I need to get this updated. On the briefing angle, you tell me what expectations/topics/angles you want covered and I will address them in a format suitable to your needs. Thank you!

  5. Merv,

    Great post – thanks for writing about our news! To clarify a few points:

    1) Aster is NOT targeting the 1TB and under market long-term. We are focused on large-scale data warehouses and companies who need more analytic power for advanced analytics/queries, even on smaller data sets. We are simply saying that for people who want to get started, we have an “Express Edition” starting at $50k for up to 1TB of data. This is in contrast with other MPP DW Appliances where the minimum price point is AT LEAST $250k, mostly because they use proprietary-engineered hardware, “bottom-heavy” (high ratio of disk to RAM/CPU cores) configurations, or (even worse) SAN’s which increase the base price for this kind of appliance. The breakthrough is Aster uses true commodity-grade hardware from Dell and can offer a compute-rich MPP DW Appliance at a much lower starting point.

    2) We have produced SQL/MR benchmarks vs standard SQL queries as part of a paper we will be presenting at the VLDB Conference in Lyon, France, later this Fall. The report is available at here: .

    Aster also has several customer case studies in the works where they’ve been able to utilize SQL/MR in ways to vastly improve queries traditional methods. ShareThis will be talking about some of this as an Aster customer at TDWI BI Executive Summit in San Diego:

    Hope this clarifies.

    1. Thanks for the comment, Steve. I think I made it pretty clear that I was talking about a specific appliance offer, but if I wasn’t clear enough, apologies to all concerned. I do make the point you will be able to upsell from there and likely want to!
      And it’s good to have another white paper to read over the weekend 😉 Thanks for doing the hard work of getting these things out into the community for discussion.
      Finally, I urge anyone who cares about these issues to check out the link to TDWI in San Diego in early August. Looks like it will be the place to be.

  6. @ Jerome:

    Aster Data Systems is a proven leader in high-performance MPP database systems for data warehousing and analytics. Aster nCluster is our single RDBMS product. We provide our nCluster software through 3 mechanisms:
    1) software only
    2) the cloud (via Amazon and AppNexus)
    3) appliances

    There was some discussion of the merits of this approach on DBMS2 last week:

    Most mature companies have more than one way to deliver their software. As a marketer, I agree that it can get confusing. The core benefits of our software remain, regardless of the delivery mechanism:

  7. Thanks, Merv. I think you were pretty clear, but there were some articles in the general press which took it as Aster going after the SMB market (not our intent), in addition to Jerome’s comments, so I thought I should clarify. 🙂 Have a great weekend.

  8. @Jerome:

    When I said “mature companies”, I was thinking of most large-scale DBMS companies who provide customers more than one way to use their software. Oracle, Microsoft, IBM, Teradata, etc. I’m not saying their technology is great/mature, I’m just saying that providing customer choice is (usually) a good thing and that companies that have been around a while wind up with multiple products to serve their customers, based on customer demand. No one questions whether they are a database company because they have more than one way to deliver the product. Sorry if my comment came off the wrong way, but I was trying to address your confusion about “where Aster is”. We are a database company with more than one product line.

    For example, Aster has a >200TB (user data) warehouse in production at MySpace – they wanted to use both Dell and HP boxes, so they just wanted the software. ShareThis, who has >10TB Aster data warehouse in production on Amazon EC2 wanted to be able to run in the Cloud.

  9. @Steve: Thanks for clearing that up. I still think having multiple delivery methods and targeting different market slices at the same time gets really expensive. Especially for a startup (which is why maybe the top dogs can afford it) such as Aster which, I understand, has deep pockets but still, it seems like a fairly “spread-thin” strategy – But hey, if you can pull it off, all the more power to you guys! 🙂

Leave a Reply

%d bloggers like this: