Database Acceleration

Faster than a speeding bullet

Databases and analytics are the core applications of many organizations in many industries. Financial, Government, Retail, Engineering, Telecoms and Web businesses all rely on fast access to large datasets.

Violin’s memory appliances provide a significant new tool for scaling the size and performance of critical database applications. Application performance increases of 10 to 100 times can be expected.

Violin memory appliances can significantly accelerate database applications of the following types

  • Transaction processing: OLTP systems rely on low latencies and high IOPS for transaction processing performance, e.g. Oracle, MySQL and SQL Serves. Violin enables high performance systems with a much smaller footprint and 80% lower power than traditional disk-based solutions.
  • Object oriented: OODBMS rely on low latency access to locks and metadata to enable high throughput processing of complex datasets by hundreds of clients. ObjectivityDB is commonly used for the largest databases. Placing the metadata on Violin enables higher throughput processing and ingestion of data with some processes accelerated by over 70 times.
  • Analytics: Data mining, data warehousing and analytics systems process TBytes of information. Oracle, MySQL, Postgress, SAS and other tools are used for these applications. Violin enables these tools to operate in real-time and can dramatically increase the query rate for some workloads.
  • Application specific: There are many applications that use embedded databases. Violin accelerates these applications by providing lower latencies and higher IOPS. For example, Violin enables Perforce (Software Configuration Management) to support 10 times the number of users.

Are you being stalled out by slow storage?

Database performance is no longer dictated by CPU performance, but by I/O performance. I/O performance issues can be reduced by using large amounts of main memory and caching, but this is an expensive approach to reducing, but not eliminating the I/O problem.  The I/O problem can be characterized as a shortage of performance in any of three areas:

IOPS:
Databases typically use 4K-16K block accesses and hence push the IOPS limits of storage. Some storage systems have been built with 10,000 hard drives to get the maximum IOPS to 1 Million and balance the CPU capability.

Latency:
A query may require many accesses to a database and its indexes. These accesses may need to be performed sequentially and hence the latency of each access is critical. Under load storage system accesses have latencies greater than 10ms.

Bandwidth:
Where the objects in a database are larger or large amounts of the database are retrieved for processing in memory, the bandwidth of the storage system is important.