Tiers & Storage

Every deployment is provisioned with a tier that determines its compute resources, a storage type that determines its disk performance, a cluster type for availability, and an availability mode for zone distribution.

Deployment Tiers

Tier	Description	Cluster Support
Free	Development and testing. Rate limited to 100 KB/s.	Standalone only
Small	Light workloads and prototypes.	Standalone only
Medium	Production workloads with moderate throughput.	Standalone only
Large	High-throughput production workloads.	Standalone and Replica
XLarge	High-capacity production with more CPU and memory headroom.	Standalone and Replica
Compute Optimized	CPU-intensive workloads - batch processing, stream transformations, analytics pipelines.	Standalone and Replica
Network Optimized	Network-intensive workloads - high fan-out, real-time streaming, low-latency delivery.	Standalone and Replica
Storage Optimized	I/O-intensive workloads - high-throughput persistent streaming with large retention.	Standalone and Replica
Ultimate	Maximum resources across all dimensions.	Standalone and Replica

Higher tiers unlock more storage options, higher network throughput, and Replica deployment support.

Network Rate Limits

Certain tiers include network rate limits:

Tier	Rate Limit	Notes
Free	100 KB/s	Always rate limited, regardless of plan
Small	1 MB/s	Basic plan only - rate limit removed once upgraded to Pro or Enterprise
Medium	10 MB/s	Basic plan only - rate limit removed once upgraded to Pro or Enterprise
Large and above	No limit	-

The Free tier is always rate limited. Small and Medium tiers are only available with rate limits on the Basic plan - once your tenant is upgraded to a higher plan, these rate limits are removed and higher tiers become available.

Resources by Tier

Tier	vCPUs	Memory	Local SSD
Free	2	1 GB	-
Small	2	2 GB	-
Medium	2	8 GB	-
Large	4	32 GB	150 GB NVMe
XLarge	8	64 GB	475 GB NVMe
Compute Optimized	16	32 GB	950 GB NVMe
Network Optimized	16	128 GB	950 GB NVMe
Storage Optimized	16	128 GB	3.75 TB NVMe
Ultimate	32	256 GB	1.9 TB NVMe

Local SSD sizes apply when using Local SSD storage type. Network storage sizes are configurable independently.

Cluster Types

Type	Nodes	Description
Standalone	1	Single node, available on all tiers. Suitable for development, staging, and cost-sensitive production.
Replica	2	Two-node cluster with synchronous replication and automatic failover. Available on Large tier and above. Requires Pro or Enterprise plan.

With a Replica cluster, if the primary node fails, the secondary takes over automatically - no manual intervention required.

Coming soon: The current 2-node Replica model will be replaced by Viewstamped Replication (VSR) - a consensus-based replication protocol that supports any number of nodes. VSR will enable flexible cluster sizes and stronger consistency guarantees.

Storage Types

Storage Type	Description	Persistence
Network Balanced	General-purpose persistent storage with balanced IOPS and throughput	Persistent
Network Optimized	Higher IOPS persistent storage for demanding workloads	Persistent
Network Extreme	Maximum IOPS and throughput persistent storage	Persistent
Local SSD	High-performance NVMe instance storage directly attached to the host	Ephemeral

Storage Availability by Tier

When creating a deployment, the storage types available depend on the tier:

Tier	Network Balanced	Network Optimized	Network Extreme	Local SSD
Free	Yes	Yes	-	-
Small	Yes	Yes	-	-
Medium	Yes	Yes	-	-
Large	Yes	Yes	-	Yes
XLarge	Yes	Yes	-	Yes
Compute Optimized	Yes	Yes	Yes	Yes
Network Optimized	Yes	Yes	Yes	Yes
Storage Optimized	Yes	Yes	Yes	Yes
Ultimate	Yes	Yes	Yes	Yes

At higher tiers you get the choice between persistent network storage (three performance levels) and raw NVMe Local SSD. Network Extreme is reserved for the Optimized and Ultimate tiers where the underlying infrastructure supports the highest IOPS.

Storage Sizes

Network storage: configurable from 20 GB up to multi-TB. IOPS and throughput are automatically tuned based on the tier.
Local SSD: size is fixed by the tier - ranges from 150 GB (Large) to 3.75 TB (Storage Optimized).

Local SSD delivers the lowest possible latency by bypassing the network storage layer entirely - ideal for ultra-low latency streaming workloads. Data on Local SSD is ephemeral (does not survive instance termination), but with Replica deployments, data is replicated across nodes, eliminating the durability concern.

Availability Modes

Mode	Description	Requirements
Single-AZ	All nodes deployed in a single availability zone. Lower cost, suitable for development and non-critical workloads.	None
Multi-AZ	Nodes distributed across multiple availability zones for high availability. Recommended for production.	Replica cluster, Pro or Enterprise plan

Multi-AZ ensures that an availability zone outage does not take down your deployment.