High Performance JVMs for Ultra Low Latency In-Memory Computing
The GridGain® in-memory computing platform accelerates and scales out data-intensive applications across a distributed, JVM-based computing architecture. GridGain solves the performance needs of companies launching digital transformation, omnichannel customer experience, Internet of Things, and similar data-intensive initiatives. GridGain is built on Apache Ignite®, which was originally contributed to the Apache Software Foundation (ASF) by GridGain Systems. Ignite is a top five ASF project and has been downloaded over five million times since the project launched in 2014.
GridGain can massively scale out to thousands of nodes and power millisecond performance for petabytes of in-memory and on-disk data. It supports multiple access APIs, including ANSI-99 SQL and key-value, and supports ACID transactions. GridGain is used for real-time transactions, hybrid transactional/analytical processing (HTAP), and high-performance digital integration hub use cases for companies in the financial services, telecommunications, software/SaaS, healthcare, transportation and logistics, and many additional industries.
The GridGain platform is ideal for digital transformation use cases. However, some low latency/high transaction volume scenarios, such as in the financial services and telecommunications industries, strain the capabilities of standard Java Virtual Machines (JVMs). GridGain caches all of the in-memory data in unbound, off-heap regions of memory. Java on-heap memory is utilized extensively by temporary objects generated by applications in runtime. Traditional JVMs may not consistently deliver the low latency required for certain transaction-intensive applications. In those specific use cases, GridGain users may need to scale out the number of nodes in their cluster in order to parallelize the workload.
The alternative to adding more cluster parallelization is Azul Zing™, a 100% Java-compatible JVM based on Oracle HotSpot. Unlike traditional JVMs, Zing decouples application performance from the amount of data kept in-memory in the Java heap. Zing is unique in its ability to provide high performance and low latency for memory-intensive applications. Zing can grow and shrink the memory heap elastically based on real-time application demands. The Azul C4 garbage collection algorithm is also able to limit JVM-related pauses to less than 30 milliseconds.
Running GridGain on Zing allows enterprises to increase the on-heap memory allocated on each GridGain node. High transaction applications generate significant Java garbage, but it is cleared efficiently by Zing. As a result, GridGain users can avoid adding extra nodes to overcome the typical JVM garbage collection challenges while maintaining consistently low latency by avoiding JVM garbage collection pauses. This allows GridGain users with applications with high read/write requirements that need low latency and low jitter to achieve their SLAs with minimal infrastructure expenditures.
For high transaction read/write applications with low latency requirements, GridGain with Zing provides major performance improvements by eliminating Java garbage collection pauses. For GridGain users, this can significantly reduce potential infrastructure costs for high transaction volume/low latency use cases by reducing the number of required GridGain nodes to achieve adequate performance for demanding use cases.