The SnappyData Platform

Built for always-on mixed workloads

OLTP, OLAP Database

  • Row and Column tables: Optimized for very fast writes, fast key/index based lookups like NoSQL and run analytic queries using vectorization and code generation techniques.

  • Collocated Processing: Optimal design for star schemas. Small/slow changing tables are replicated across the cluster and large tables as partitioned but related data sets are collocated based on relationships. So, now Join processing requires little to no shuffling of data.

  • Distributed transactions: Achieve ACID transactions across distributed tables.

  • Elastic Clustering: Data servers are connected to each other in a true p2p cluster ensuring group consensus in the presence of failures. The cluster can be elastically scaled at run time.

  • Shared nothing persistence: While tables can be purely in-memory with redundant copies in the cluster, any table can also be reliably persisted to local disk. Disk technology uses continuous appends completely avoiding disk seek delays.

  • Not Only SQL: Store JSON, Java/Scala Objects, nested or self describing data.

  • Enterprise class security, backups

High Availability & Disaster Recovery

  • In-memory redundancy: Data is replicated with synchronous consistency within a cluster, or across racks for HA. Data can be asynchronously replicated across data centers for DR. When replicated asynchronously, data is always eventually consistent.

  • Shared Nothing Parallel Persistence: Data can be persisted on disk and each replica maintains its own independent copy on disk.

  • Active-active WAN replication:

Deploy on-premise or any cloud

  • Run SnappyData on-premise using commodity servers
  • Or, as a Amazon/Azure Cloud service
  • Or, Docker containers

SnappyData is also available as a Spark package and accessible from any Spark application as a dependency.

Explore SnappyData's Vision

SnappyData's ACM SIGMOD 2016 paper is available