How in-memory computing will change the way we obtain information in real time
We have reached the fifth article in our series “The Revolution in management systems caused by Omniera” and now we are going to talk about in-memory computing and its advantages for obtaining real-time information.
In previous articles in the series we have talked about the principal transformations caused by this revolution, the advent of the collaborative economy, how Big Data can be applied in Omniera business and the impact of the Internet of Things on company management.
In-memory computing is an architecture that transfers databases to memory, giving much faster access to large quantities of information. In-memory computing has become important primarily because it does away with the need for disk I/O – I/O is the input and output, reading and storing of information in hard disks.
With in-memory computing, there is also no need for ACID database transactions. A database transaction is a sequence of operations executed with a single logical unit of work. A logical unit of work, in turn, must have four properties, designated by the initials ACID – Atomicity, Consistency, Isolation and Durability, in order to qualify as a transaction.
By eliminating the need for an ACID transaction when using OLTP (Online Transaction Processing) applications, in-memory computing makes it possible, for instance, for a large mass of data stored in the cloud to be processed very quickly.
To make in-memory computing possible, database management systems (DBMS) have been transformed, giving rise to the IMDB or In-Memory Database. IMDB technology has been identified as the answer to problems of database performance, since it is able to load and execute all the data in a large database in memory. This does away with a substantial number of inputs and outputs (I/O) which hamper transactions and create performance problems for database systems.
In-memory computing has only become feasible with the constant improvements in hardware architecture. Thirty years ago, companies wanting to install an ERP system could only do so with the help of vast mainframe computers with a huge processing cost which put them out of reach of most companies. With the invention of the microchip and major changes in hardware and software architecture, ERP systems became more attractive.
Changes in the data storage model
According to the survey The Internet Economy 25 years after .com, by Robert Atkinson, published by the Information Technology and Innovation Foundation in March 2010, storing one gigabyte on a hard disk cost around 44 dollars in the year 2000. By 2010 the cost had fallen to a mere seven cents of a dollar.
The world’s first mass-marketed computer, the IBM 1401, launched in 1959, was 1.5 meters high and 90 cm wide, and held 4,096 bits of memory. At the time it could perform 193,000 calculations with 8-digit numbers in a minute.
This computer was leased out at a cost of 30,000 dollars per year. In 2012, however, the world’s cheapest computer, the Raspberry Pi, cost 25.34 dollars. Today’s smartphones which fit in your pocket are connected to the internet via mobile data or wifi networks, cost a few hundred dollars and have storage and processing capacities far superior to the CRAY-1A supercomputer launched by Cray Research in the 1970s. The CRAY-1A weighed around five tons and cost something in the region of nine million dollars.
ERP Systems with In-memory technology
Companies that have been using the same ERP system for a number of years are familiar with the difficulties in managing large databases and extracting historical data and trend analyses from them in the traditional way.
For decades ERP systems suppliers created overnight routines, background routines and routines operating on non-business days to provide strategic information for companies when a large amount of data had to be processed.
When a company needed information not included in these routines there were always difficulties. As a result, the success of organizations depended increasingly on performance in finding and extracting information and analyses of this sort.
Now some ERP systems providers have combined cloud computing techniques and in-memory computing to offer products with hardware, storage, operating systems, management software and in-memory data search resources, with the data for processing being held in the RAM instead of being read from disks or flash storage, thus improving performance.
The trend is for this new generation of architecture to be applied to the majority of ERP systems ERP, and within a few years many companies will start benefiting from this alternative with major gains in data search performance.
In order to support the dynamics and diversification of different types of business, companies adopting or migrating to an ERP system will need to find a solution based on an innovative, robust, scalable and functionally comprehensive platform that enables them to manage all the channels – an omnichannel solution combining the possibility of cloud computing with the speed offered by in-memory computing.<
Since 2013, a number of companies have been using ERP systems with this format and capability. Although ERP systems using in-memory computing are not yet a majority, it is clear that this is the most rational solution for companies.
Migration to Omniera-compatible ERP systems
Another interesting trend we are seeing now, and which can be expected to strengthen in the years ahead, is the migration from traditional ERP systems and legacies, whose architecture is not suitable for handling large databases, to Omniera-compatible ERP systems which are enabled for in-memory computing.
Before the possibility of uploading and keeping large databases in memory existed, many companies had to perform an incredible amount of juggling, including deleting data to keep the performance of their ERP systems at reasonable levels. Others invested a veritable fortune in hardware, but since the ERP architecture was based on transactional databases, with the need for disk I/O and few options for improving performance, these investments were not enough for the needs of their businesses and did not improve operating performance sufficiently.
On the other hand, in-memory ERP solutions have a high analytical potential and high-performance architecture, which means that users will be able to search large volumes of data in great detail and in real time, thus offering them a robust, secure and agile structure which at the same time will simplify their internal processes and assist in decision-making at all levels of the organization
In a recent case, a clothing (lingerie) company opted for an omnichannel ERP with in-memory computing. Now it can complete in seconds procedures for which its old ERP system needed four hours – proof of the gains in performance that these systems can offer.