Chat iconGet in touch

Why next gen BSS needs an in-memory database

By October 15, 2014 No Comments

For sure, VoltDb meets all the pragmatic needs of virtualised cloud optimised systems and Network Function Virtualization. Unlike legacy databases, it can scale elastically and is not tied to physical hardware concepts such as the number of CPUs or servers. It provides horizontal scalability, high availability with no single-point of failure, uses logical and physical separation of redundant resources and a shared-nothing architecture. Compared to legacy databases, it is easier to deploy and maintain, easy to automate and does not require specialist DBA personnel. On the Capex side, it does not require expensive fibre channel infrastructure and SAN storage. All of the above deliver on the promise of reduced costs and faster time to market today. 

However, it is the performance of VoltDB in-memory database that will be of critical importance in the future.  Network operators are already faced by a data explosion due to the massive increase in smart phones and the wide range of mobile based product and services from video content to health services to music services to security services (and many more expected).  This data growth is soon to be compounded by the increasing connectivity demanded by internet of Things (IoT) and wearables with Gartner predicting there will be nearly 26 billion IoT devices by 2020.  Network operators need to handle all of this data in Real time.

Traditional database systems such as disk-based RDBMS are plagued with processing overheads and are simply too slow to ingest data, analyse it in real-time, and make decisions.  As Michael Stonebraker et al reported in the seminal paper for the Association for Computing Machinery’s Special Interest Group “OLTP through the Looking Glass, and What We Found”. This explained, the need to carry Index Management, Write-Ahead Logging, Locking, latching and Buffer management, means traditional databases were only spending 12% of processing on useful work with the remaining allocated to housekeeping. Recent benchmarks indicate this can even be as low as 4% useful work. These techniques evolved originally for data integrity, but now the overhead prevents traditional databases from scaling to meet contemporary data volumes and workloads.

VoltDB in-memory database eliminates all of these legacy DBMS Overheads.  Data and associated processing is partitioned together and distributed across the virtual nodes in a shared-nothing cluster. Data is held in memory for maximum throughput (no buffer management needed), and each single-threaded partition operates autonomously (no need for locking and latching). This delivers orders of magnitude better performance than a conventional DBMS and without comprising data consistency and integrity ensuring ACID (Atomicity, Consistency, Isolation, and Durability) transactions.

Openet’s very low latency real time systems seamlessly utilising VoltDB’s in-memory capabilities delivers a limitless and linear ability to scale cost effectively so as to meet the evolving real-time needs of the evolving network.