Cloudflare has announced Workers KV Instant, a new operational mode for its key-value storage system that leverages the company's internal Quicksilver storage architecture. Currently available in private beta, the service is designed specifically for infrequently updated global application configuration data and high-frequency read operations.
Sub-Two-Millisecond Latency and Rapid Replication
According to Cloudflare, Workers KV Instant achieves 99th-percentile (p99) read latencies under two milliseconds, which is over 100 times faster than classic Workers KV reads. 95th-percentile read speeds reach microsecond levels. Global edge replication is also accelerated, pushing 99 percent of write updates across more than 300 edge locations in approximately 250 milliseconds without requiring explicit time-to-live (TTL) cache expirations.
API Compatibility and Specific Operational Differences
Developers interface with KV Instant using standard Workers KV API calls, including get(), put(), list(), and delete(). However, the mode introduces specific operational changes: developers must explicitly set the mode attribute to instant when creating namespaces, metadata capabilities are omitted with getWithMetadata returning null, and list operations return all matching keys without pagination support. Write operations are restricted to one per second per namespace.
Divergent Pricing Model for High-Read Workloads
Pricing for KV Instant favors read-intensive architectures. Class B read operations cost $0.20 per million requested keys, representing a 60 percent reduction compared to classic Workers KV pricing. In contrast, write and metadata operations (Class A) are priced higher at $0.10 per operation, and storage costs $100 per megabyte per month. Namespaces are limited to a maximum of 1,000,000 bytes in total size and up to 10,000 key-value pairs, with individual keys capped at 300 bytes.
AI-generated editorial illustration; not a documentary photograph.
Source: Introducing Workers KV Instant — powered by Quicksilver

Leave a Reply