Hazelcast IMDG Hot Restart

With Hazelcast IMDG® High-Density Memory Store, enterprise users are able to use a Hazelcast® cluster as primary source of records by creating large in-memory caches to store large volumes of data.

Since memory stores are volatile space, data is lost when the memory stores are recreated after a server restart –either during a planned shutdown or as a result of an unplanned outage. Hydrating such large caches then becomes highly challenging. In some cases, it may take anywhere from several hours to a few days to build and then load the full data into the cache.

Hot Restart Store

Hazelcast Hot Restart Store is a high performance feature that provides persistence to enable reloading of data into the cluster after a restart. The Hot Restart feature stores the states of cluster members along with the data stored in their caches, on disk in a format that is purpose-designed for ultra high-performance data reload.

Hazelcast Hot Restart Store is a high performance feature that provides persistence to enable reloading of data into the cluster after a restart. The Hot Restart feature stores the states of cluster members along with the data stored in their caches, on disk in a format that is purpose-designed for ultra high-performance data reload.

Hot Restart Store Architecture

Hot Restart Store Architecture

Hot Restart Store can be used with IMap, Web Sessions, Hibernate and ICache. The data that gets stored in HD store of supported data structures, also gets persisted in its local Hot Restart Store. In the event of a cluster restart, data is reloaded into HD memory from the Hot Restart store.

Highlights

  • Ultra high-performance persistence store, optimized for SSD for fast data reload
    • 1 TB of data can be loaded in 76 seconds on a 10 node cluster
    • 500k writes per second per server node
  • Linearly scalable – each node has its own independent store
  • Data entirely loaded into RAM on reload – standard Hazelcast in-memory speeds
  • Supports partial restart to allow to restart a cluster with an incomplete member set
  • Hot backup via Management Center: snapshot of Hot Restart Store at a certain point in time – helpful to perform testing, QA or reproduce an issue on production data
  • Handles real world operational challenges – moving data to another server/cluster, allowing different servers to take part in the restart phase etc.
  • Optimised for SSD howeve the log structued storage approach also works well with rotational drives (HDD)

When to Use Hot Restart

With Hazelcast High-Density Memory Store, you can store so much data in cache that it can take a very long time to load that all up in the event that the entire cluster needs to be restarted, or you have a lights out event in the data center or cloud region.

With Hot Restart, you can reload this data very fast and within operational SLAs. Seconds rather than days.

One final note: Hazelcast is not designed to be a System of Record on its own. It acts as an operational store or cache, with the system of record being in an RDBMS or NoSQL. These systems provide much richer data management functionality that is intentionally not designed into Hazelcast. Using Hazelcast together with a System of Record gives the optimal combination of speed, scale and rich data management functions.

Products in this Use Case:

Hazelcast IMDG

Learn More

Next Steps:

Contact Us

Hazelcast.com

Menu