In-Memory Database Systems

In-Memory Database


An In-Memory Database Systems (IMDS) is a database management system that primarily relies on the main memory of the computer (RAM) for data storage compared to a traditional database management system which employs a disk storage mechanism. They are a growing subset of a database management system (DBMS) emerged in response to new application goals, system requirements, and operating environments.

In-Memory Database Systems (IMDS) are faster than disk-based databases since the internal optimization algorithms are more straightforward and execute fewer CPU instructions. Accessing data from the RAM (main memory) eliminates seek-time when querying the data, which provides faster and more predictable performance than disk. In theory, IMDS scales well beyond the terabyte size range.

The key reasons behind concerning ‘‘data storage mechanism” as the hearth of enterprise software systems is it is the most significant part of the software that determines how quick an application responds to a request, and the loss of data is mostly unacceptable since the critical business operations.

Key Categories/Type of In-Memory Database Systems

We can categorize in-memory databases systems into four critical classes according to different optimizations they have adapted.

Key-Value Store: In a Key-Value Store we store the data as key-value pairs like the data structure ‘‘hash table” where keys are used to retrieve the data. Example: Redis & Memcached

Document Store: In a Document Store we store the data in collections that contain key-value pairs which encapsulate key-value pairs in JSON (JavaScript Object Notation) or JSON like documents. In a document store since values are not opaque to the system, we can query the data by values as well as keys. Example: MongoDB

Column Family Store: In a Column Family Store the data is stored as a set of rows and columns where we group columns according to the relationship between the data. Column family shares the characteristic of being schema-free with document stores. However, the implementation is very different. Example: Cassandra

Graph Database: In a Graph Database or Graph DBMS we represent data in the form of graph-structured as nodes and edges. It is easy and flexible to process data in this form. Example: Neo4j

Storage Optimization for In-Memory Database Systems

Recent evolution in RAM sizes and rapidly dropping market prices of RAM in the past years has inspired considerable increases in the prevalence of database systems that keep the entire database in memory. Due to this in-memory computing became economic among a wide variety of applications. Though it’s economically feasible to use the main memory as the primary data store still RAM is a limited resource and expensive compared to a disk.

A primary goal of recent research works around IMDB is to improve in-memory storage optimization. The freer memory, the larger systems to be stored in the database, which enhances the performance and the cost efficiency. In the In-Memory Database Systems research, recent works focus on approaches to identify hot and cold data so that we can store the hot data in the RAM and move the raw(cold) data to a cheaper disk store.

A traditional RDBMS relies on ACID (Atomicity, Consistency, Isolation, Durability) consistency model that ensures all the transactions are correctly committed and do not the corrupt database, and the data are consistent. Whereas an IMDB relies on BASE (Basically Available, Soft-state, Eventually Consistent) consistency model to achieve scalability, high availability, and high performance.

Advantages of In-Memory Database Systems

Some of the critical benefits of IMDS are

1) IMDS is used to cache countless amounts of data continuously. Using this cached data, we can ensure speedy response times for searches.

2) The ability of IMDS to store session data, allowing for the customization of live sessions and guaranteeing optimum website performance.

3)  IMDS is used to process events for improved complex event processing.

Applications of In-Memory Database Systems

Some of the critical demands of IMDS are

1)   IMDS is used widely in algorithmic trading and some other applications used in financial markets to provide instant manipulation of data and to identify and leverage market opportunity.

2)    Various B2C web applications like e-commerce and social networking sites also use IMDS to cache a portion of their on-disk database.

3)    IMDS is also used widely in real-time embedded systems or real-time operating systems (RTOSs) like IP network routing, telecom switching, and industrial control.

2 comments On In-Memory Database Systems

Leave a reply:

Your email address will not be published.

Site Footer