We have an application that uses Bucket4j for throttling with Hazelcast as the backing in-memory data grid cache. Since Hazelcast is no longer supported as of Pega Infinity '25, we are looking for throttling on a low-latency caching solution that works across multiple (preferably on) specified node types. The release notes at Platform '25 release notes mention Pega provides built-in alternatives. How can we use these built-in alternatives?
Hi @DennisA69 , there isn’t a like-for-like distributed in-memory cache to re-point Bucket4j at. Everything Hazelcast used to do is now handled by Pega natively (Kafka-backed cache sync, node-local caching with DB-backed coordination, Kubernetes for node discovery), but it doesn’t mean that Pega provides these capabilities to the outside consumer natively.
Probably the closest platform-native replacement for a custom token bucket is a dedicated Queue Processor. You can control throughput by limiting consumers/threads and pinning it to a specific node type — this gives you a predictable, cluster-wide TPS ceiling without any shared in-memory counter.
Another advantage it gives you is more compatibility with Pega Cloud, which can make migration much easier.