Intelligent Tiered Storage

Overview

In a traditional deployment, all data grid entries are stored in JVM heaps to provide the fastest performance possible. However, as the repository of data grows and grows, the following issues can arise:

  • Cost - While RAM performs much better than a hard drive, it is also much more expensive. Using GigaSpaces' MemoryXtend module to store hot data in RAM and warm data in SSD provides great business value. Additionally, the PMEM storage option provides performance similar to RAM, but at a cost closer to that of SSD.
  • Garbage Collection - The bigger the JVM heaps get, the harder the garbage collector works. Using MemoryXtend to store and manually manage some of the data off-heap (i.e. in the native heap instead of the managed JVM heap) allows using a smaller JVM heap and relieves the pressure on the garbage collector.
  • Access - As data ages from hot to warm, it resides in relatively fast-access memory. However, cold data is generally moved to a data lake where querying the data can take much longer. Using AnalyticsXtreme to partition the cold data into frequently accessed and infrequently accessed tiers enables rapid retrieval of needed information, while the rest of the data is safely archived.

Using GigaSpaces' MemoryXtend and AnalyticsXtreme modules, you can implement intelligent tiering for the full life cycle of the data. Hot data stays in RAM for fastest access time, while MemoryXtend manages the storage of warm data in high-performance memory such as SDD. AnalyticsXtreme ensures that the transition from warm to cold data is smooth and precise, as the data is moved from costly memory to a data lake (file system or object store). In the data lake, customers can differentiate between frequently accessed cold data and infrequently accessed archive data using AnalyticsXtreme's batch indexing feature.

This data tiering ensures that your organization has the optimal balance between cost and performance.