How it Works
How it Works
Big Data combines three key elements: Volume of data. Velocity for which data changes, is created, or can be accessed. Variety of sources from which data stems.
Any of these elements break a traditional RDBMS. Spire addresses each of these challenges, plus the power of real-time SQL.
Built on Hadoop – for massive volume
Spire is built on Hadoop, the most scalable and reliable foundation for Big Data. Spire is 100% distributed, easily handling any volume of data and any number of users.
Hadoop is a great foundation providing fully distributed data storage. In addition, Spire stores data in its own version of HBase optimized for high throughput, low latency, and durability under load. Leveraging the fundamentals of the BigTable and F1 models, Spire handles the distribution of the data, making it simple to scale out across hundreds of servers. Providing cluster management at a cluster level, not at a server level, creates operational efficiencies that cut costs down for administration and infrastructure management.
Built for velocity with two-layer optimization, distributed indexing, and the Spire VQM
Velocity at scale is often the most difficult of the three elements. Spire enables consistent performance with a two-layer (Global and Local) optimization approach. The Global Optimization Engine pushes down computation only to nodes with relevant data. On every node, the Spire Virtual Query Machine (VQM) enables Local Optimization, considering memory, network latency and other factors, thus consistently ensuring the best throughput and performance for the query. Without local optimization, computations can get stalled on over-taxed nodes, not leveraging the full resource pool and bringing the cluster to a halt.
Fast reads and writes, while under load, are integral to an operational data store. Spire enables true real-time Big Data applications through distributed indexing, producing results in milliseconds, against even the largest data-sets, supporting many concurrent events and users.
Spire also has no garbage collection pauses, resulting in more reliable performance at the 99th percentile.
ANSI SQL and Joins to access a variety of data types
Spire is the next generation data unification platform, allowing consolidation of data from disparate sources, creating a common schema accessible from multiple applications for multiple use cases. Today’s enterprises understand the value of accessing more data or bringing offline data live to reduce costs or improve their knowledge of their customers. Spire is built against the ANSI SQL specification and tested with TPC-H and other standard benchmarks.
Consolidated data often requires the use of joins to make sense. Joins are very hard in a distributed system. Spire’s Global and Local Optimization engines make joins possible for BIg Data. This a fundamentally new approach to solving this complex problem. With a complete view of the data and the ability to slice and combine it, applications can be built with more accuracy and greater insights.