Answering more and more complex queries is a challenge for today's Business Intelligence, the same as analyzing sales trends is a challenge for managers. No one doubts in the point of business analysis, nonetheless all involved people are still looking for better and better solutions. For quite a lot of time, OLAP (OnLine Analytical Processing) has been the right choice. However, the very beginnings of OLAP are dated back even to 1962, and - even though the development processes were being continued, minor changes never were enough.
Thereupon, there finally was a time to put the in-memory OLAP idea into effect.
Traditional OLAP (MOLAP, ROLAP, and Hybrids)
The idea of OLAP (or MOLAP, as both these terms refer to as the same) could be boiled down to presenting data in a way that enables watching it from different perspectives. As the point is to watch all the aspects of each situation, people used to 'describe' situations with elements called OLAP cubes.
Think about spreadsheets. These are nothing more than the ancestors of today's cubes. Even though they're being used constantly, are incomparably less usable than cubes because of their two-dimensionality. That's why the OLAP cubes were considered revolutionary - they allowed people to highlight three dimensions of data at the same time.
The primary form of online analytical processing, called OLAP or MOLAP (because it's exactly 'Multidimensional OLAP) is characterized by no use of relational database. Data is being stored in dedicated array storages, which are multidimensional. As a consequence, there is processing needed - if the cube is expected to be filled with data, there is pre-computation required, what slows down the whole analyzing. However, we'll consider the time factor later in this article.
Unlike usual OLAP, ROLAP (Relational OnLine Analytical Processing) bases on data kept in databases. All in all, everything could be boiled down to tables, there. Base data with dimension tables are stored as relational tables, while the aggregated information also require another tables. The thing is there always has to be additional parameter attached to every information. It's 'where' the information is stored. The consequence is growing complexity of SQL statements.
The third type of OLAP, so called HOLAP (Hybrid OnLine Analytical Processing) is much more difficult to be defined. The problem is specialists do not agree what exactly the base of HOLAP is. The point - data needs to be divided between relational databases and specialized storages.
Basically, the three stated above are the most common types of OLAP methods. However, some people point out WOLAP (Web-based), DOLAP (Desktop), and RTOLAP (Real-Time) as well, nonetheless the range of their uses is much tighter than in case of the three.
The most meaningful OLAP vendors are: Microsoft, Hyperion, Cognos, Business Objects, MicroStrategy, and SAP AG. And if you kept this list in mind, you would happen to discover something interesting as soon as we turn to in-memory OLAP.
OLAP In-memory revolution
The idea of OLAP is commonly appreciated. If it wasn't, the biggest worldwide IT giants wouldn't be interested in fighting for market shares. However, traditional OLAP is not as good as it should and as some people supposed it to be in terms of efficiency and architecture. As a result of gathering the benefits of traditional OLAP with in-memory technology the in-memory OLAP was introduced. Why is that? What features are considered the most important? Finally, is the in-memory OLAP innovative enough to replace former, traditional type and repeat its success?
Time needed for analysis
One of the most meaningful disadvantages of traditional OLAP was a relatively long time needed for performing business analysis.
Once everything is being operated within the memory, the time needed for analyzing is now significantly shorter. In practice, all relevant data is loaded to memory, what makes connecting with external data sources no longer needed.
No pre-calculation limits
Traditional OLAP was devoted to regular pre-calculations being the way for supporting the analysis. Thank to in-memory OLAP, with calculations being performed on the fly, there is no need to wait any longer. Moreover, limited number of operable pre-calculations no longer matters, as well.
Inexpensive RAM memory
In-memory OLAP technology had been known for years, nonetheless formerly it wasn't possible to introduce it into practice because of the price of RAM. Nowadays, RAM is relatively cheap what makes it more accessible.
64 bit architecture
Nowadays the majority of the systems are based on 64 bit architecture which replaced the 32-bit. Thank to that, more and more operations can be operated at the same time. Furthermore, their complexity also no longer is so meaningful (systems become more and more tolerant to complex queries).
In-memory OLAP software
Basically the traditional OLAP is provided by the biggest worldwide companies like IBM, Microsoft and SAP. They have managed
to develop a great and complex structure for OLAP users, what was time and money-consuming. Thereupon, no one should be surprised with the
fact they're not the first that turned to a new technology.
That is why in-memory OLAP has been developed mostly by smaller vendors, like Qliktech, Applix, Hiqube, Tibco Spotfire, Panorama Software, Information Builders and others.
Even though the in-memory OLAP still is quite a new to market, there are more and more different platforms using this method. Which one is the best or - at least - worth considering? Up to now, there are a few main "players" in the market and their products - Qlicktech QlikView, IBM Cognos TM1 (formerly Applix), and Palo - deserve attention. However, new solutions are being worked out constantly, therefore one should be prepared for sudden changes. Being still a new idea, in-memory OLAP requires a lot of improvements, and a lot still is left to be done. Who will be the first to develop fully efficient solution? Whose platform will attract the lion's share of the market? Next years will show. Nonetheless, currently there are three pretenders.
QlikView by QlikTech company is a self supporting platform which provides a bunch of up to date capabilities. Understand they're constrained on fighting with former in-memory Business Intelligence solutions popularity, authors of QlikTech decided to prepare a platform answering most of current needs, when analyzing with iPad or BlackBerry is a standard. However, what truly is important are consolidation capabilities. It actually doesn't matter what source your data comes from, with QlikView you'd be able to analyze it anyway. By advanced searching capabilities, QlikView users can quickly get to the most interesting areas of data - it's not a problem to show all data. Much more difficult it is to display only what truly is important. And that's what QlikView is very well prepared for. Finally, efficient results visualizing is easier than ever before - with QlikTech's platform you can choose from differentiated graphs, tables, and charts. It's obvious the designers did their best to allow users to get results in the clearest way possible, what - by the way - means something different for each user.
What else matters, is the speed. With QlikView you can reach the totally new level of analyzing timeliness, even in case of the most complex queries. Thank to phenomenal scalability supported with multi-core 64-bit processors, the platform can be used by the largest and the tiniest companies equally well.
IBM Cognos TM1
IBM missing any of greatest IT innovations? That's just impossible.
Thereupon, after acquiring Applix, IBM Cognos is now the widest known "player" in the new market. Nonetheless, a little exaggeration it would be to presume that TM1 - IBM's in-memory OLAP platform - automatically is the best. It's good, but if the best, it depends.
Like every other modern platform, TM1 works with 64-bit architecture - it's just the standard while thinking about performance. As a result, TM1 has no problem with analyzing even complex queries - that's most important, because, as we all well remember, the need for in-memory OLAP itself has arisen from low speed of analyzing using traditional methods. However, what designers paid most of attention to was not only speed, but also a wide range of supported business uses. Thereupon, TM1 is supplied with enterprise planning software, personal scenario tools, and integrated reporting and scorecarding. Moreover, by integration with Cognos BI platform, TM1 supports the newest trends in Business Intelligence, providing facilitated drive-based planning and roll forecasting.
Finally, to ensure that each customer gets exactly what he needs, TM1 enables triple methods for viewing (interfaces). Cognos TM1 Web, Cognos Contributor client, and Microsoft Excel - all up to each one's own habits and preferences. It's also worth mentioning that Cognos Express - the BI platform for mid-sized companies is based on TM1 architecture.
Palo OLAP Server
What should the best platform look like? Maybe like Palo, prepared by specialists from Jedox who paid a lot of attention to making their solution as user friendly as possible. The way each manager gathers his company is specific and different in every case. This is why it's truly difficult to define one method acceptable for all of them, however Palo seems to have solved this problem. With data organized in regular cubes, elements, cubes, and attributes, Palo OLAP Server operates fast and efficiently allowing managers to focus on what truly is important. Once measured, the results were up to hundred times better than in case of traditional analytical processing. That's pretty impressive, though.
Then, why should
people choose Palo instead of any other in-memory OLAP platform? Because
it's simple, multidimensional, and full of additional capabilities.
What requires pointing out is real time data aggregation and calculating - it's understood that analyzing is tightly tied to planning as well. Furthermore, modeling - which formerly was complicated and problematical - has also got simplified (even if your primary tool is MS Excel).
What else? Multiple interfaces (plus additional and modifiable extensions) allow users to find the what suits them best. Numerous dimensions. Plenty of features and capabilities that make Palo OLAP Server another great figure in in-memory OLAP platforms palette.
In memory OLAP Resources
Stefan Walthers private blog about QlikView & Business-Intelligence. The blog is a good knowledge base on Qlikview, it is organized into such categories as Qlikview tips...