Since the advent of relational databases for business applications, the data stored in RDBMS has been utilized for computing critical optimizations like Manufacturing Resource Planning (MRP), fraudulent credit card transactions etc... The earlier databases relied on storage that was very expensive and hence the data was minimized; for e.g. saving data in tables without the century digits, which later led to the Y2K problem. It involved significant cost and effort to solve the Y2K problem which involved changing millions of lines of COBOL code with very little documentation to upgrading firmware in embedded devices. As storage became cheaper, the constraints on size of data being stored were eliminated and databases started storing non text data too, such as images and documents.
Once the data is available in the database, processing the data involves frequent retrieval / storage which involves significant I/O, making the processing slower and leading to a new problem of Big Data. A MRP run for a large manufacturing organization could take many hours, and determining a fraudulent transaction could take hours after the transaction has been completed. The migration of servers from 32 bit processor to 64 bit processor technically provides up to 16 exabytes of Random Access Memory (RAM). The price of DRAM memory chips are dropping every year. Due to these factors, RAM is the preferred choice for storing the entire data temporarily for computing instead of doing frequent I/O to disk. “Memory is the new disk”, says NY Times.
The systems that create the orders for a MRP are done using Online Transaction Processing (OLTP) Systems. The Online Analytical Processing (OLAP) Systems take the OLTP data and provides insights into various KPIs to provide a dashboard view to take critical business decisions. The OLAP transactions crunch massive data and takes time to arrive at these dashboards. These are done as a post-mortem activity after the transactions have been created in the OLTP systems.
When architected correctly, the data kept in-memory can speed up computing up to 1000x and OLAP activities can be performed as a real time transaction where OLTP and the OLAP can be combined instead of doing the OLAP as a post-mortem activity. There is a paradigm shift happening wherein when in-memory data is used, online MRP activities can be performed whenever orders are raised or fraudulent transaction can be detected as soon as a credit card is swiped.
In-Memory Computing (IMC) is the storage of the entire data required for processing in the main Random Access Memory (RAM) of dedicated servers rather than in relational databases operating on comparatively slow disk drives.
Ramco as part of their M.U.S.I.C initiative have ventured into M - Mobility (Mobile First), U – User Interfaces (Uber cool UI for Gen-Y), S – Social, I – In Memory, and C – Context Aware. Ramco’s Minnal framework is an In-Memory Computing framework which can solve difficult problems in the area of Advanced Planning, Scheduling and Optimization combined with Execution Control using Minnal multi-agent technology which can scale to work with terabytes of data. Minnal augments various Ramco products like ERP, HCM, GRP, MRO to run MRP, payroll, e-governance batches, component maintenance planning etc…