IBM solidDB Universal Cache

Features and benefits of IBM solidDB Universal Cache
IBM solidDB Universal Cache is the industry’s first relational, in-memory
database caching technology that accelerates virtually all leading relational databases, increasing their performance up to ten times. It uses the familiar SQL language to enable applications to achieve tens of thousands of transactions per second with response times measured in microseconds. By combining the relational, in-memory data management capabilities of solidDB Universal Cache with the versatility of disk-based databases, applications benefit from the best of both worlds.

Highlights

Accelerates Virtually All Industry-Leading Relational Databases with Industry’s First Universal Cache
solidDB Universal Cache complements IBM DB2 and Informix, Microsoft SQL Server, Oracle and Sybase databases to accelerate access to their data. Its two major components are relational, in-memory database software that delivers performance-critical data with extreme speed; and high performance data synchronization software that copies data back and forth between the in-memory cache and a specified backend database. As a result of the broad support for industry-leading backend databases offered by solidDB Universal Cache, application teams can leverage the same skills and experience to accelerate performance across a variety of database technologies and projects.

Achieves Extreme Speed with In-Memory Database Technology
solidDB delivers extreme speed because it keeps data in main memory at all times rather than on disk. Applications can take advantage of its capability through standard ODBC, JDBC, SQL interfaces. solidDB works under the premise that all data to be accessed will be in main memory, all the time. Thus, it has data structures and access methods specifically designed for storing, searching, and processing data in main memory, with efficient concurrency control mechanisms. This has two performance advantages over conventional, disk-based databases. First, solidDB eliminates the need to transfer data blocks from disk to main memory. This is because any data requested by the application is already in main memory. Second, solidDB is faster than disk-based databases, even if the latter cache all data in main memory, since solidDB employs data structures and access methods that are optimized for main memory access. Extreme speed is further achieved because solidDB Universal Cache can be co-located and linked with one or more applications for increased performance and reduced response times. Similarly, multiple applications can be linked to a dynamic library comprised of the complete query processing code. The applications may access the common memory-resident data via a shared memory segment. The shared memory access allows multiple applications to share a single solidDB instance. This avoids network latency and the overhead of inter-process communication, as well as context switching in the operating system.

Adapts to Different Application and Deployment Needs
solidDB Universal Cache can be deployed as a read-only cache or as a read-write cache. When configured as read-write cache there are two possible deployment options.

When configured as a read-only cache and the data is owned by the backend database. This means the data stored in solidDB Universal Cache can not be modified by the application. In this configuration, applications can modify the data directly in the backend database and changes can be synchronized to the in-memory cache, transaction by transaction, automatically or on-demand. This configuration is ideal for applications that require fast access to data that changes occasionally, such as price lists, or reference or lookup data.

When configured as a read-write cache and the data is owned by the cache, applications can read, add, modify, or delete data in the in-memory cache, but not in the backend database. Changes are propagated from the in-memory cache to the backend database, transaction by transaction, automatically, or on-demand. This configuration is useful for applications that have stringent service level agreements that demand short response times, for a variety of data intensive operations.

Finally, when configured as a read-write cache and the data ownership is shared; applications can update the same data in both the in-memory cache as well as in the backend database at the same time. In this case, changes to the data can be automatically propagated in both directions. Conflicts are detected and resolved by using pre-defined conflict resolution methods. This configuration is especially useful when applications need to update the data in the backend database while the data is also cached for read-write access.

Using a dynamic and flexible SQL passthrough capability, applications can access both cached and non-cached data, negating the need to maintain an explicit connection to the back-end database. This pass-through feature checks if the SQL statement can be handled locally by the cached tables, or if it must be redirected to the back-end database. The SQL passthrough capability can be enabled at a session or a transaction level and can be changed dynamically at runtime, thus allowing the application a wide degree of flexibility and greater level of control.

Additionally, with flexible and configurable data aging, solidDB Universal Cache optimizes the amount of memory allocated to cache and ensures the data closest to the application is the most relevant and active.

Provides Data Durability and Persistence
solidDB Universal Cache implements an ACID compliant (atomicity, consistency, isolation, and durability) in-memory cache, that persists data to disk using checkpointing and transaction logging. This is especially useful in read-write cache configurations (where the cache owns the data), or in the event of loss of network connectivity between the in-memory cache and the backend database. In the latter case, the in-memory cache continues to process the workload, persists the changes to local disk, and accumulates the changes for synchronization with the backend database at a later time, without data loss.

Provides High Availability and Instant Failover Between solidDB Universal Cache Instances
solidDB Universal Cache can deploy in a two-node, hot-standby configuration, that maintains two copies of data synchronized between two cache instances. In case of system failure, failover between cache instances can occur in less than a second. This feature helps to protect users from unplanned outages. Whether a system failure or a planned "switchover" between the pair of cache instances, the secondary instance is designed to quickly take over the processing load, preventing data loss. The failover is engineered to be transparent to the application; solidDB Universal Cache's ODBC and JDBC drivers maintain the database connection and session attributes.