Aerospike: An Open Source, Flash-Optimized, In-memory, NOSQL Database

SI Team
Thursday, December 11, 2014
SI Team
Aerospike, the first distributed database built to take advantage of flash storage, represents the next generation of caching technologies, in-memory technologies and NoSQL databases. It delivers RAM like speed with the economics of flash or solid state drives (SSDs). It is often used to replace memcache, redis deployments that need to scale without having to deal with sharding and cassandra deployments that need predictable performance with significantly smaller, more cost effective clusters.

Aerospike was founded to make database technology previously only available within the walls of Internet giants like google, available to all developers. The company was founded in 2009, first customer deployments went live in 2010, and the company open sourced it's proven technology in 2014.

Rapid Scaling
In a short amount of time, Aerospike has been recognized as a leader and a visionary by industry analysts. It used as a cache at e-commerce and retail companies like SnapDeal, Williams Sonoma and Kayak and Aerospike has become the database of choice in real-time bidding - new internet infrastructure - for holding an auction to determine who can place an ad in front of a consumer. This is an eco-system of internet wide platform companies like AppNexus, Inmobi, Chango, eXelate and others that are personalizing consumer experiences across web, mobile, video, search, gaming, social media and other properties.
These platforms operate at millions of transactions per second and process tens of terabytes of data in order to track big data or user context -every click, swipe and user action on the internet - in order to understand what the user wants and respond with relevant and personalized offers in milli-seconds. The entire real-time bidding process, from the time a user clicks on a web page to the time that the page is painted, along with personalized messages, offers or ads, is about 100 milliseconds. This means that every entity in the process - the publisher or supply side platform, the ad network, the ad exchange, the demand side platform and data management platforms - has just 5-10 milliseconds in which to determine what data they have and decide whether to bid, how much to bid and if the auction is won, what offer to put in front of a consumer.

Modern Stack
Companies using big data to sense and respond are adopting a new stack - first, a tier of stateless web application servers that scale up and down based on consumer demand. Next, Aerospike, the front edge, operational database, used as a server side cookie store or session store, for device id mapping and user context. It also contains additional data processed in and loaded from the third tier - relational databases or analytic systems like Hadoop or Vertica.

Share on Twitter
Share on LinkedIn
Share on facebook