It is near the end of the year. 22nd December 2010. It’s been snowing heavily here in the UK, which is rare. As year end is approaching I’d like to write about the current trend on DWBI. Reflect on what has happened in 2010, and what might take off in 2011-2012.
Before I start I’d like to remind the reader that this is what I see. So it’s not the entire truth. It’s only a fragment of the market i.e. towards where I live (UK) and the products I’m familiar with. If you are a DWBI vendor, I’m sure you see more a complete picture than me in terms of your products. So please correct me.
Trends in DWBI Software
DWBI Suite: As a complete DWBI tool set/suite, by far MSBI is the most widely used (number of companies using it, not revenue). OBIEE, SAS, BO & Cognos follow at a far distance. Cognos is not a suite, where’s the ETL tool? It’s more appropriate calling it “IBM suite”, where they bundle InfoSphere DataStage and DB2 into the mix. There was an ETL tool in Cognos 8 BI (called Data Integration), but in V10 it’s gone. SAS is not a complete DWBI suite as it has no DW. Their ETL is DI Studio. BO has an ETL tool (BODI), but no DW/DB either.
ETL: terms of features, Informatica 9 is at the top, but SSIS is the most widely used. By far. Others are Data Stage (InfoSphere), OWB (now ODI in 11g), Ab Initio, BODI. For a more comprehensive list (of ETL tools) see my old article here. Some of my colleagues who have used Ab Initio claimed that it is the fastest ETL tool due to in-memory look up/pinning and parallel operations and that it scales well. But I doubt it. Their approach is very secretive. We had to sign an NDA before seeing the presentation. This usually means that there is nothing behind the wall. It’s a marketing technique where if you are low it’s best to keep yourself hidden. ODI & BODI are about equal at number 3 behind SSIS & Informatica (in terms of usage). There has been a change of perception since 2009 about SSIS, that it is no longer an inferior product like DTS. And the market take up since 2008 has been amazing. There are a lot more SSIS job than Informatica. Here in London many insurance companies use SSIS as their ETL tool, where as banks use more Informatica, mainly because of historical reasons. One or two banks still use Hummingbird Genio. I admit the new ODI has very good features (in particular the Knowledge Module), which could position it above SSIS (in terms of features, not usage).
Reporting tool: unlike 8 years ago, companies are no longer buying a stand alone reporting product like Crystal Report any more. They buy a suite. Lots of companies use SSRS because it comes “free” with SQL Server and the IT dept personnel are able to use it (because they can download & install it – unlike BO & Cognos). Companies with more budget usually opted for BO, Cognos & MicroStrategy. BO XIR2 developer edition is £1750 per user whereas Webi is £500 per user. The Dev Ed allows you to create reports, whereas the webi is to read/run reports. Compare that price with SSRS which is free. Which is why some companies are migrating from BO to RS (I’ve come across 3 companies in London this year). With regards to prices I want to point out that prices vary widely from companies to companies. Not only BO but all vendors. Company A could pay £1000/user whereas company B only £200/user for the same product.
BPM (Business Performance Management), aka EPM (Enterprise PM) or CPM (Corporate PM): Hyperion (Oracle), SAS, Board, BO & Cognos. Board is semi Self Service BI. BO’s one is called Planning and Consolidation. In 2007 SAP bought OutlookSoft and Pilot but they seems to be gone now. Hyperion is by far the top in this market. Pity OutlookSoft is no longer there to challenge Hyperion. SAS looks good and has proper Financial Mgt.
BPM definition is never clear, lots of marketing & labelling. Budget vs actual, planning, KPI, all thrown into the mix. Traditionally it’s Financial Mgt, like Hyperion & SAS but then all sorts of BI are thrown in, including dashboard, scorecard & KPI. So when people say B/E/CPM, clarify first what they mean by that.
OLAP: SSAS is as strongest as ever. By far the most widely used cube tool. I would estimate AS is probably 60% of the market now (usage, not revenue, i.e. out of 100 companies who use cubes, 60 of them use SSAS – my estimate, not market research). I’ve seen a few banks now use SSAS. Oracle Essbase & IBM PowerPlay follows far behind. Oracle 11g OLAP option could be a contender in the next 2 years. MicroStrategy is a respectable player in OLAP and has been in the market for a very long time (20 years?). Consistent with ROLAP since the beginning. Very useful to read their comparison to other vendor here. There is a big change going on with SSAS in terms of UDM & in-memory OLAP. Read MS BISM (Semantic Model) here (TK Anand), here (Chris Webb), here (Boyan Penev), here (Teo Lachev), here (Marco Russo), here (Jason Thomas) and here (Thomas Ivarsson).
RDBMS (SMP): the most widely used RDBMS for DW is without a doubt SQL Server (usage, not revenue). By far. Then Oracle, then DB2. Oracle and especially DB2 are used only by big companies, whereas SQL Server is used by companies of all sizes. Sybase & Informix are not in the game. Just like MySQL & Ingres, I only heard like once a year (if that) where people use Sybase or Informix for data warehousing. There has been a total change of perception with regards to SQL Server capacity handle big DW. 5 years ago the perception (especially amongst London banks) was that SQL Server can’t handle big DW (say upwards of 3 TB). So they used Oracle/DB2 for “big DWs” and SQL Server is approved for “small DWs”. The perception now is that SQL Server 2008 capability for large DW is equal to Oracle 11g. I know an investment bank that uses SQL 2008 for their huge DW (upwards 15 TB). This take up is largely due to SQL 2008’s partitioning capability. Both Oracle and DB2 are still very respectable for DW platform. Some people still think that they are “slightly better” than SQL Server, in terms of performance for >5 TB DW.
MPP: I think Teradata & Netezza are still leading (again, number of companies using it not revenue). Exadata is gaining popularity, but it’s kind of 1/4 MPP. Definitely not full MPP. Not even half. Read here (vs Twinfin) and here (Teradata). Good if you are thinking transactional, but for MPP DW it’s a question mark. Seems that PDW & NeoView are not in the game. This year I heard their promotions/marketing but not their implementations. Even Greenplum is more used than PDW & NeoView (London, not sure about US & other countries). In my opinion if you use MS BI (SSIS, SSAS, SSRS) and are choosing an MPP, it is better to choose Teradata than PDW. We have OLEDB for Teradata, which enables IS, RS & AS to connect to Teradata better. Read MS-TD integration here and here. I was disappointed to hear that PDW had technical problems and that the take up is not good. When I wrote this I was very optimistic about PDW. Kognitio WX2 is MPP, but using commodity blades.
Visualisation: What’s that? OLAP client? Reporting client? Charting? Well, all of the above. MicroStrategy, CubePlayer, Tableau, QlikView, Excel, SSRS, SCA (Strategy Companion Analyzer), Panorama, Crescent. For SSAS client, I think Crescent would be the winner, but that’s 12 months away. How about now? From features, not Panorama, but Tableau, CubePlayer and SCA. From architecture/installation, etc, not Excel, but SCA. Vidas Matelis composed a comprehensive list of SSAS client here. The issue with SSAS client is not the features, but the vendor. Majority are small vendors. Which ones will still stand in 10 years time? How can we be sure that they will not be bought and then killed like ProClarity?
Trend in DWBI Jobs
It is important for DWBI architects and managers to consider the resource/people/skill factor when choosing DWBI software. Employees & contractor also need to actively look at the trend, to consider what training they need to do in 2011. Agencies know this well, but employees/contractor usually not: best site for salary/rate and trend is ITJobsWatch. Below I only mention the rate (contract), for salary just click the “Permanent” tab on the IT Jobs Watch site.
ETL: Informatica rate is higher than SSIS. In London market (not UK rates), SSIS current market rate is £350-450 DOE, where as Informatica is £500-600 (because you can’t download the product like SSIS). Note that the £450 SSIS also comes with SSRS & SQL Server skills and to a degree SSAS, so it’s a clear advantage to the recruiter.
OBIEE: is gaining popularity. More people wants OBIEE Developer. This is caused by version 11g. Good rate too, OBIEE 11g was £500 in Nov but now £600. Big shortage (many installations but few developers). Unlike MS BI which is saturated.
MPP: No change in Teradata & Netezza skills. Still very low volume. Stable/no change in rates, 375-450. Risky to get into because of low volume. And for most people, it’s impossible to get into. Not only you can’t download evaluation version, but where can you get hardware from? The only chance is taking training at Marylebone (Teradata) or Marble Arch (Netezza).
Solvency II: Solvency II is the name of the game now, driving lots of DWBI projects. So people with Lloyds market knowledge has advantage. And this trend is going until 2012, perhaps 2013. Again it’s London (and Europe, e.g. Dublin, Brussels, Zurich) but it does not happen in for example Boston or Melbourne.
BO: BO is interesting: decreasing volume (companies are moving to RS) but increasing rate (400-450): see here. The best of BO is a chance to get SAP experience.
QlikView: skill is more in demand compared to last year (number of jobs doubled) but rates decrease (understandably): was 400 now 350. Again this is London rate, outside London is lower, see here.
Future Trend in BI (not DW but BI)
Some people ask me what’s next in BI. This was my usual answer. There are 2 things that will gain popularity in the next 2 years:
• Self Service BI. They already gain popularity. Users want to compose their own reports. Connect to the DW / DM themselves. I can imagine it will be a mess in a few years, but there you go, that’s the trend. Main stream BI will still be built by IT, only “personal” view are in the users’ hands. Main players are QlikView & PowerPivot. And Board to a degree. No body heard Rosslyn yet (“What’s that?” is the response I’m getting at the moment). In terms of features, QlikView is still perceived to be much better than PowerPivot as PowerPivot is still version 1. But the take up is good. I’ve seen a few companies using it so it’s only a matter of time before PowerPivot overtakes QlikView.
• Streaming BI. In weather and science research streaming BI is old news. Temperature analysis for example. But in commercial companies this is new. Stream Insight is the major player (part of R2). I think it will take off in the next 2 years. People wants to analyse the data as soon as it comes in. FX trading in an example (well, other asset classes too). Any kind of online investment trading really. O yes and risk analytics. It’s kind of not stream at the moment, but quite a big flow. It’s kind of against all DW principles that’s why it’s a bit odd. DW is batch. Nightly. Not second by second streaming. It’s a revolution, I know. But it’s coming. Read here for an intro and here for architecture.
People make mistakes, so if you know anything is incorrect please rectify me. As usual I welcome discussion & questions at email@example.com.
Vincent Rainardi, 22/12/2010