Skip to main content
A cluster profile is a preconfigured performance and cost setting that determines how a data streaming cluster behaves based on its underlying storage, latency expectations, and workload characteristics. Each profile packages the right combination of infrastructure choices—such as disk-based storage for low-latency access or object storage for cost-efficient retention—to deliver a predictable experience without requiring users to tune individual parameters. By selecting a cluster profile, you can align your cluster with the needs of your applications, whether you prioritize real-time responsiveness, throughput, or cost optimization. All cluster profiles run on the URSA engine, StreamNative Cloud’s unified data streaming engine. The profile you pick determines the write-ahead log (WAL), metadata store, and supported protocols; the cluster type (Pulsar or Kafka) determines the API surface.

Cluster profile types

StreamNative Cloud offers two cluster profiles. Pick the one that matches your workload’s latency and cost priorities.

Latency-Optimized profile

The Latency-Optimized cluster profile is designed for real-time, interactive, and mission-critical workloads that require consistently fast data access and low end-to-end latency. Backed by high-performance disk storage, this profile delivers predictable, single-digit to tens-of-milliseconds latency for producers and consumers, making it ideal for use cases such as event processing, fraud detection, user activity tracking, and any application where responsiveness directly impacts user experience or business outcomes. By prioritizing speed over cost, the Latency-Optimized profile ensures smooth, high-throughput performance even under demanding conditions.
Latency-Optimized cluster profile architecture
Clusters using the Latency-Optimized profile are expected to deliver end-to-end latencies in the 5–200 millisecond range.
The infrastructure underpinning the Latency-Optimized profile varies by cluster type:
  • Pulsar Clusters use Apache BookKeeper as the low-latency WAL. The metadata store is ZooKeeper by default; Oxia is available on request. Pulsar Clusters support the native Pulsar protocol, and the Kafka protocol is available through the KSN protocol handler with full Kafka feature parity. Lakehouse integration (Iceberg, Delta) is built in via the URSA storage layer.
  • Kafka Clusters use KRaft and local disks as the low-latency WAL, delivering native Apache Kafka with the full Kafka feature set. Lakehouse integration is built in.

Cost-Optimized profile

The Cost-Optimized cluster profile is engineered for throughput-intensive and cost-sensitive workloads that benefit from large-scale, durable, and economical object storage. While this profile accepts higher access latencies—typically in the hundreds of milliseconds—it provides significant cost savings and unlimited scalability, making it a strong fit for data pipelines, analytics sinks, long-term retention, and applications that do not require ultra-fast consumption. By leveraging object storage, the Cost-Optimized profile offers an efficient balance of price, durability, and flexibility for workloads where latency tolerance is higher and storage economics matter most.
Cost-Optimized cluster profile architecture
Clusters using the Cost-Optimized profile are expected to deliver end-to-end latencies above 200 milliseconds.
The infrastructure underpinning the Cost-Optimized profile uses object storage—such as Amazon S3, Google Cloud Storage, or Azure Blob Storage—as the WAL, and Oxia as the metadata store. The capabilities exposed vary by cluster type:
  • Pulsar Clusters currently expose the Kafka-compatible protocol only on the Cost-Optimized profile. Native Pulsar protocol support for this profile is coming after the Apache Pulsar 5.0 release. Lakehouse integration is built in.
  • Kafka Clusters expose the native Kafka protocol with lakehouse integration built in. Kafka transactions and topic compaction on the Cost-Optimized profile are coming soon.

Pulsar Clusters

Pulsar Clusters support all three deployment options. Serverless automatically manages performance and cost characteristics without user-defined profiles.
On Pulsar Clusters, the Cost-Optimized profile currently supports the Kafka-compatible protocol only. Native Pulsar protocol support for this profile is coming after the Apache Pulsar 5.0 release.

Supported deployment options

Deployment optionLatency-OptimizedCost-Optimized
ServerlessN/AN/A
DedicatedSupportedNot supported (Coming soon)
Bring Your Own Cloud (BYOC)SupportedSupported*
* Cost-Optimized BYOC Pulsar Clusters currently support the Kafka-compatible protocol only. Native Pulsar protocol support is coming after the Apache Pulsar 5.0 release.

Supported features by deployment type

Cluster profiles do not apply to Serverless deployments. Serverless automatically manages performance and cost characteristics without user-defined profiles.
CategoryFeatures
ProtocolPulsar (native), Kafka (via KSN, with full Kafka feature parity)
Table formatsDelta, Iceberg
CatalogUnity Catalog, Snowflake Horizon Catalog, S3 Tables, Google BigLake
StorageDisk (internal only)
Availability ZoneSingle AZ (internal only)

Kafka Clusters

Kafka Clusters are currently in Public Preview. For a comparison with Kafka compatibility on Pulsar Clusters (KSN), see Kafka Cluster vs. KSN.
Kafka transactions and topic compaction on the Cost-Optimized profile are coming soon. Use the Latency-Optimized profile today if your workloads require these features.
Kafka Clusters run native Apache Kafka on the URSA engine and support both cluster profiles. Serverless support is planned.

Supported deployment options

Deployment optionLatency-OptimizedCost-Optimized
ServerlessComing soonComing soon
DedicatedSupportedSupported
Bring Your Own Cloud (BYOC)SupportedSupported

Supported features by deployment type

Serverless support for Kafka Clusters is coming soon.

Next