In-memory computing: An emerging step-change in enterprise application architecture
Author: Andrew Powell, General Manager – Commercial
At Sapphire Orlando this year I was presented with a concept that has really got me thinking because it is such a game changer. For the 15 years I’ve been focusing on Enterprise Application architecture all our assumptions were based on the concept of relational databases with the separation of storage across permanent stores (like disk) and RAM. That’s why the OLTP (online transactional processing) and OLAP (online analytical processing/data warehousing) was originally separated- to spread the loads of these relational databases which are physically constrained by the speed limitations of reading and writing to disk. We all talked about the virtues of this architecture for years, i.e. separation of OLAP and OLTP and the n-tier architecture.
In-memory computing changes this. Imagine if all the necessary enterprise data was ‘in-memory’ (i.e. RAM). In simple terms there is no need to spread the load across separate OLTP and OLAP databases in that scenario. The mind boggles at where this is heading but I did hear from one SAP representative that their emerging SAAS ERP offering (called ‘By-Design’) would be 100% in-memory by the end of the year. In fact the ‘By-Design’ solution is an insight into this new way of thinking because it combines transactional and analytical functions into the same screen. For example – a sales order entry screen also presents an up-to-date sales chart in the same screen.
In terms of what is happening today, there are a number of SAP customers using Business Accelerator Blades which, in simple English, allow for sub-sets of analytical data to be stored in-memory. Complex queries are done in seconds instead of hours and Google type searching is possible. Under this scenario the disk based storage is not eliminated but the way disk is used, in relation to RAM, and the relative mix of the two is clearly shifting. It will be fascinating to watch this development evolve over the next 5 years.
For those who find this interesting I’d guide you to a rather technical website for the “Hasso Plattner Institut” http://epic.hpi.uni-potsdam.de/Home/InMemoryDataMgmt. Also there is a simple summary in this interview of Hasso Plattner http://sapphirenow.blogs-sap.com/2010/05/06/hasso-on-hasso/
