Meta unveiled ZGateway, a stateless proxy layer for its distributed key-value store ZippyDB, designed to solve connection and reliability problems caused by more than one million client hosts. The gateway now processes more than 1 billion operations per second and handles roughly 40% of ZippyDB traffic, according to a report published by InfoQ in September 2026. The architecture reduces total persistent connections by approximately 19 times, Meta's modeling shows.
ZippyDB powers product metadata, counters, configuration, and other workloads across Meta's globally distributed infrastructure. Under the old direct access model, clients connected straight to database hosts serving the shards they needed. A single client could touch tens of thousands of shards spread across hundreds of thousands of database hosts, creating what Meta described as a dense many-to-many connection mesh. The company's modeling estimates that per-host connection counts drop by approximately 97 to 98% with the new architecture. In a controlled test above 90% CPU utilization, six of roughly 1,350 tenant buckets shed traffic while the remaining tenants processed 99.9% of requests without rejection.
The report states that connection storms could contribute to file descriptor exhaustion and out-of-memory conditions under the previous system. Meta said ZGateway places a managed proxy tier between clients and ZippyDB servers, with clients maintaining sticky connections to regional gateway hosts while database servers receive connections only from the controlled gateway fleet. According to the company, the gateway can authenticate and authorize requests, apply per-tenant admission control, resolve shards, use local caching, and batch or coalesce requests before forwarding them to ZServer replicas. Because the gateway sees traffic from multiple clients, it can combine requests that individual client libraries can't observe across processes.
The additional network hop represents a deliberate tradeoff, the report explains. Md Shuvo, an architect at Sherbrook, noted in a LinkedIn post that "adding an extra network hop actually improves overall latency by freeing up database nodes from the brutal overhead of connection management." ZGateway provides a centralized location for load balancing, cross-region resilience, caching, and overload protection—capabilities that weren't possible when clients connected directly to database hosts. The gateway runs as regional tiers discovered through ServiceRouter, Meta's hyperscale service mesh solution, keeping clients geographically near their gateway. Meta uses its existing C++ ZippyDB client as the gateway's request engine and supports both a pure proxy and read-through cache tier, allowing the architecture to handle batching and coalescing that individual processes couldn't manage alone.
Meta plans to route all ZippyDB traffic through ZGateway while exploring agent-operated controls, selective co-location with ZServer, and a multi-process architecture for stronger fault isolation. The company can progressively route traffic by service, shard prefix, percentage, or region, with a global kill switch available for rollback. The proxy layer shifts where shared traffic management occurs—moving authentication, admission control, and request batching from individual clients to a managed tier that sees traffic patterns across the entire system. For distributed systems managing massive connection volumes, the proxy pattern offers a path to centralize control without requiring clients to coordinate directly, though each additional layer introduces latency that must be weighed against operational gains.