There is a growing trend towards the adoption of in-memory computing technology and with this comes some confusion with regards to what it’s all about and the terminology surrounding it. In-memory computing integrates innovative technology with business processes and applications and helps businesses to experience sustainable competitive advantages in terms of performance, IT architecture simplification, process innovation and performance.
What is in-memory computing?
In-memory computing (IMC) moves data traditionally stored on hard discs into memory. Rather than storing information in a relationship database, it stores it in the random access memory (RAM) of the dedicated servers.
Random Access Memory (RAM) is random access because the location of the stored information doesn’t affect the access speed. The more RAM that’s available, the more applications can run at the same time without slowing down the system.
Storing information in RAM dramatically reduces latency and speeds up the process of data analysis. Latency is the amount of time it takes for the memory to respond to a command. Generally, the lower the delay or latency, the faster the device. At Gigaspaces, you can find out more about an in-memory data grid.
A relational database organizes data into tables that are linked based on data common to each. The main benefit is to create meaningful information by linking them.
Transactional and analytical data
Transactional and analytical data processing is possible with an IMC database. IMC databases are specialized databases that move data into memory. They allow for intelligent data processing and close to real-time results to complex data queries.
Transactional data is the data captured from business transactions, such as when a product is bought or sold. Analytical data supports decision-making and historical data stored in a data warehouse makes this possible. IMC enables the running of operational applications and data analysis in close to real-time on a single database.
A traditional data warehouse versus an IMC-enabled data warehouse
In a traditional data warehouse that stores raw data, part of the data is extracted to data marts. Business intelligence applications can request results from the data mart for processing and visualizing of final results.
When raw data is stored in an IMC-enabled data warehouse, business intelligence applications don’t request partial results from data marts. They get results straight from the IMC-enabled data warehouse. By querying the built-in calculation engine, results are close to real-time and the data mart layer is obsolete. Frequent updating of raw data means that transactional applications can feed data directly into the IMC-enabled data warehouse.
The value of in-memory computing
It is possible to waste the power of IMC or for applications to make inefficient use of it and prevent it from offering a sustainable competitive advantage. A simple speeding up of processes is possible but talent is required to identify the specific information needs of a business. Effective management of data volumes and respective transactional and analytic processes is necessary.
The gain in performance offers the potential for innovative applications that differ from those of competitors. For example, the manufacturing industry could consider complex scheduling based on the most recent data or the finance industry could analyze near real-time current risk exposure.
With data marts being obsolete, the complexity of data models is significantly reduced. Data models have the advantages of less testing, easier adaption, faster creation and less potential restart points. With less complexity comes fewer possibilities for error.
The near real-time calculation of analytics from raw data offers flexibility in terms of the integration of additional data sources and modification of the analysis. New data sources are easy to plugin as an additional source of information as every calculation begins from raw data. It doesn’t require much effort to change analytical procedures that make up the analysis, as this only requires a change to a query.
With the flexibility and simplification, this can potentially reduce the total cost of ownership (TCO). There is less effort to develop new data models and to maintain and develop existing data models.
It is difficult to give a definition of Enterprise Architecture (EA) as this is rather elusive. One definition of EA is “the structure of components, their inter-relationships and the principles and guidelines governing their design and evolution over time.” To offer the most value, IMC must properly fit into enterprise architecture. Enterprise architecture includes an understanding of how value creation translates to applications, information technology and data models. If enterprise architecture is already set up, the process is easier.
The role of enterprise architecture in an organization is cross-disciplinary and requires integrating diverse methods, skills and tools. Business role players, the IT team, and the EA team need to collaborate on an ongoing basis.
Steps to implement IMC
The first step is to define the capabilities that provide the most business value. The next step involves identifying changes required to enterprise architecture encompassing technology, processes and data models.
The third step is to decide on a migration plan that either offers a smooth transition with quick wins or a big bang. For example, replacing a traditional data warehouse with an IMC-based data warehouse gives quick performance gains. However, if the long-term goal is to reduce the TCO by simplifying and achieving maximum performance gains, this means optimizing applications and data models for IMC.
In-Memory Computing does not deliver durable value if it is not implemented mindfully. An unfocused collection and analytical evaluation of data is not likely to provide the necessary insights. When looking at digital marketing trends in 2021, businesses are eagerly searching for the type of business insights that give them a competitive advantage.
Hopefully, the above information clarifies in-memory computing and some of the terminology associated with it. Making the most of In-memory computing requires a clear strategy that includes identifying and evaluating use cases and managing implementation. If organizations follow a clear and well worked-out strategy, it offers a great potential in terms of cost reduction, performance, process innovation, flexibility and simplification.