What is Varnish?
The HTTP acceleration engine at the center of your delivery stack.
The engine that keeps your origin out of the equation.
Varnish is an HTTP accelerator and reverse proxy that serves data at near-memory speed, shielding your infrastructure from cacheable traffic before it hits your origin.
When a request arrives for a cached object, Varnish responds in under half a millisecond. The origin only gets involved when it has to: when the cached object has expired, or when the content is not cacheable. At scale, that distinction is the difference between stable infrastructure and infrastructure that buckles.
For two decades, the world's most demanding web properties, streaming platforms, and enterprise pipelines have relied on Varnish to sit between their users and their backends. The same engine that handles a global broadcaster's live stream also accelerates AI data pipelines, Kubernetes routing, and software supply chains.
What can Varnish accelerate?
A lot of things. The same engine handles:
Web & API delivery
Dynamic applications, CMS platforms, and high-concurrency API traffic.
Video & media streaming
Linear live streams, VoD titles, and high-volume asset delivery.
AI & data infrastructure
Object storage and GPU training pipelines at exabyte scale.
Software supply chain acceleration
Public and private registries including Docker Hub, npm, PyPI, JFrog Artifactory, and more.
Open source or Enterprise?
Varnish Software continues to actively develop, maintain, and evolve Varnish Cache as an open-source distribution built to accelerate any type of HTTP workload and offload infrastructure.
Varnish Cache
A caching foundation providing HTTP acceleration and reverse-proxy functionality. It includes major core engine features and packages a comprehensive set of extra modules that aren’t available out of the box in a Vinyl Cache setup, including:
- In-process TLS termination
- Official container images & Helm chart
- Dynamic backends with TLS to origin
Varnish Enterprise
Engineered for high-concurrency workloads and massive data footprints, extending the core engine with specialized storage, security, and edge compute capabilities.
Enterprise Capabilities
Storage & Routing
- MSE 4 persistent cache
- High availability, replication & clustering
- Traffic routing
- Tag-based invalidation
Security & Compliance
- Perimeter WAF & JWT auth
- Software supply chain security
- Distributed rate limiting
Programmability & Ops
- 50+ native enterprise modules
- WASM edge compute
- Varnish Controller UI
Whether deploying community open-source instances or powering critical infrastructure, Varnish remains the definitive engine for serving data fast while shielding origins.
How Varnish fits into your stack
Varnish Enterprise sits as a unified delivery layer between your data sources and your end consumers. The architecture adapts to your environment rather than requiring you to adapt to it.
Web app backends
Application APIs, legacy CMS platforms, and dynamic applications.
Video & media origins
Linear streams, VoD titles, and high-volume assets.
Object storage / S3
Exabyte-scale cloud storage, data lakes, and AI datasets.
Software repositories
Public and private registries: Docker Hub, npm, PyPI, JFrog, and more.
Varnish Enterprise
One extensible platform
Multi-tier architecture moving data at near-memory speeds.
Caching & acceleration
Sub-millisecond cache lookups, request coalescing, resilient hybrid object storage (memory + disk).
Clustering, HA & load balancing
Shared cache clustering, active-active replication, origin shielding, L7 backend directors
Perimeter security
In-process TLS, JWT authentication, edge WAF, supply chain filters.
Programmable edge compute
Atomic VCL routing, per-request header control, cache policy definition.
Observability & telemetry
Logging and metrics engine with real-time filtering. Multiple log formats (JSON, NCSA). Metrics and traces via Prometheus and OTel.
Global portals & APIs
Web users, banking API clients, and retail consumers.
OTT streaming devices
Smart TV platforms, connected devices, and live media audiences.
AI & compute clusters
GPU training arrays, HPC simulation nodes, and analytical systems.
Varnish use cases for critical infrastructure
Open source heritage and enterprise support
Open source heritage
Varnish has its roots in open source, and that heritage shapes how the commercial product is built and maintained. Varnish Enterprise runs on the same open-source core, with our own patches layered on top, so VCL code written for Varnish Cache runs on Enterprise too. Verified binaries, signed packages, visible release pipelines, and predictable upgrade paths are core parts of the engineering process.
Support & stability
Commercial deployments come with SLA-backed vendor support, validated software streams, and upgrade pathways that don't require a rearchitect. Storage failures isolate without restart penalties. Clustered deployments absorb node changes without traffic disruption.
Battle tested
When infrastructure teams ask whether they can depend on Varnish for critical production workloads, 20 years of deployments at global scale is the answer.
Engineering principles
What makes Varnish fast?
Infrastructure depends on more than a standard proxy layer. These are the engineering decisions that make Varnish behave differently under load.
CORE ARCHITECTUREVarnish is built around storage performance and handling massive amounts of concurrent requests. Every request follows a rapid hash lookup straight into system memory, shielding origin infrastructure from traffic. Concurrent requests for the same uncached or expired content are coalesced into a single upstream fetch. Cache-to-cache clustering means a peer absorbs the fetch if a node misses, so the backend never takes unnecessary load.
STORAGE ENGINEMSE manages terabyte-class caching without exhausting system resources. Objects are stored in large, pre-allocated files with their own internal filesystem-like behavior, removing overhead from the OS filesystem. Cache persists across restarts, so a reboot or upgrade doesn't mean starting cold.
Hybrid memory and disk architecture
Pins high-demand hot objects inside RAM while cold data automatically falls back to low-cost disk arrays under a self-regulating Memory Governor.
Persistent and resilient by design
Cache persists to disk, so a restart rebuilds from persistent storage instead of starting cold. If a disk fails, you lose only the portion of the cache stored on it, not the entire cache.
Varnish Configuration Language (VCL) is a domain-specific programming language designed to manage per-request behavior programmatically.
VCL compiles to native code via C and swaps atomically at runtime without dropping a single active connection.
TLS termination runs directly inside the multi-threaded cache runtime rather than through an external proxy hop.
Eliminates internal proxy-hop latency to hit benchmarks clear of 150+ Gbps.
MODULARITYSupports over 50 built-in Varnish Modules (VMODs) to control complex request logic and enterprise security controls right at the perimeter.
Varnish Enterprise core capabilities
An extended capability breakdown:
| 01
Storage architecture and delivery performance
|
|
| 02
Edge security engineering and secure transport
|
|
| 03
High availability, invalidation, and edge compute
|
|
| 04
Enterprise observability and integrations
|
|
Open source tooling for Kubernetes and sandboxing
Gateway API
A full Kubernetes Gateway API implementation. Delivers standards-compliant routing with Varnish's performance.
tinyKVM
Designed for high-performance computing, tinyKVM allows standard Linux binaries to run securely in micro-VMs with near-native execution speeds.
Get started
Where to go next
Running serious traffic?
Talk to the team about Varnish Enterprise.
Deploying on Kubernetes?
Install the Gateway API implementation directly into your cluster.
Building a CDN or streaming platform?
Explore Ora Streaming and Varnish CDN.
Try open source
Spin up Varnish Cache using official project container configurations.