Upstash vs Redis Cloud vs Dragonfly: Best Redis Alternative (2026)

Redis is everywhere — caching, sessions, queues, rate limiting, pub/sub. But the Redis licensing change in 2024 reshuffled the market. Upstash offers serverless Redis, Redis Cloud is the official managed service, and Dragonfly is the high-performance open-source alternative.

Quick Comparison

Feature	Upstash	Redis Cloud	Dragonfly
Type	Serverless Redis	Managed Redis	Self-hosted (or cloud)
Protocol	Redis-compatible	Redis native	Redis-compatible
License	Proprietary (service)	RSALv2 + SSPLv1	BSL 1.1 → Apache 2.0
Pricing	Per-request	Per GB RAM	Self-host (free)
Free tier	10K commands/day	30MB database	Free (self-host)
Serverless	Yes (core feature)	No (provisioned)	No
Global replication	Yes (multi-region)	Yes	Manual (self-managed)
Max memory	Unlimited (disk-backed)	Up to TBs	Limited by server RAM
REST API	Yes (HTTP)	No (TCP only)	No (TCP only)
Best for	Serverless/edge	Traditional Redis workloads	High-performance self-hosted

Upstash: Serverless Redis

Upstash is Redis reimagined for the serverless era. Pay per request, not per hour. No server to manage. Works from edge functions, serverless platforms, and anywhere with HTTP.

Strengths

Serverless pricing. Pay per command, not per provisioned instance. Idle databases cost nothing. Perfect for variable workloads and side projects.

HTTP/REST API. Access Redis from edge functions, Cloudflare Workers, Vercel Edge, and any environment where TCP connections aren't available. This is uniquely powerful.

// Works from edge functions (no TCP needed)
import { Redis } from '@upstash/redis'

const redis = new Redis({
  url: process.env.UPSTASH_REDIS_REST_URL,
  token: process.env.UPSTASH_REDIS_REST_TOKEN,
})

await redis.set('key', 'value')
const value = await redis.get('key')

Global replication. Multi-region read replicas for low-latency access worldwide. Reads go to the nearest replica.

Additional products. QStash (serverless messaging/queues), Upstash Kafka, and Upstash Vector — all serverless, pay-per-use.

Generous free tier. 10K commands/day free. Enough for development and small projects.

Weaknesses

• Higher latency than in-memory Redis (HTTP overhead + network)
• Not all Redis commands supported (most are, but some advanced ones are missing)
• Cost can exceed managed Redis at high volumes (millions of commands/day)
• Not suitable for ultra-low-latency use cases (<1ms)
• Disk-backed storage means performance profile differs from pure in-memory Redis

Best For

Serverless applications, edge computing, Vercel/Cloudflare deployments, and any project where pay-per-use is more efficient than provisioned instances.

Redis Cloud: The Official Service

Redis Cloud is Redis Ltd.'s managed service. It runs actual Redis with all features, extensions (Redis Stack modules), and enterprise support.

Strengths

Full Redis. Every Redis command, every module (RediSearch, RedisJSON, RedisGraph, RedisTimeSeries). No compatibility concerns.

Performance. In-memory, provisioned instances with consistent sub-millisecond latency. Best raw performance of the three options.

Active-Active geo-replication. Multi-region read AND write with conflict-free replicated data types (CRDTs). Critical for global applications needing multi-region writes.

Redis Stack. Built-in modules for search (RediSearch), JSON documents (RedisJSON), time series, graph, and probabilistic data structures.

Enterprise support. SLAs, dedicated support, compliance certifications (SOC 2, HIPAA, PCI-DSS).

Weaknesses

• Pricing by provisioned RAM. You pay for capacity whether you use it or not. Expensive for variable workloads.
• License change concerns. Redis's RSALv2/SSPL license change in 2024 made some organizations uncomfortable.
• No HTTP API. TCP connections only. Doesn't work from edge functions or serverless environments without connection pooling.
• Minimum costs. Even the smallest paid plan costs more than Upstash for low-usage scenarios.
• Overkill for simple caching. The full Redis Stack may be more than you need.

Best For

Traditional applications needing full Redis compatibility, sub-millisecond latency, and advanced modules. Enterprise deployments with compliance requirements.

Dragonfly: High-Performance Self-Hosted

Dragonfly is a Redis-compatible in-memory data store that claims 25x higher throughput than Redis on a single instance. It's designed for modern multi-core hardware.

Strengths

Performance. Multi-threaded architecture (Redis is single-threaded). Dragonfly uses all CPU cores, delivering 25x throughput on multi-core machines. Single instance can replace a Redis cluster.

Memory efficiency. Uses less memory than Redis for the same dataset. Efficient memory allocation and data structure implementation.

Redis compatibility. Supports 200+ Redis commands. Drop-in replacement for most Redis use cases. Existing Redis clients work without changes.

Snapshots without blocking. Dragonfly's snapshotting doesn't pause the server — a significant improvement over Redis's BGSAVE.

Open-source. BSL 1.1 license (converts to Apache 2.0 after 36 months). Self-host freely.

Weaknesses

• Self-hosted only (mostly). No managed cloud service with the maturity of Redis Cloud or the simplicity of Upstash. Dragonfly Cloud is emerging.
• Smaller ecosystem. Fewer tutorials, fewer community resources, less battle-tested in production.
• Not all Redis commands. Most are supported, but some edge cases and modules are missing.
• No Redis Stack modules. RediSearch, RedisJSON, etc. are not available. Need alternatives.
• BSL license. Not as permissive as Apache/MIT. Cloud providers can't offer it as a service (by design).
• Operational burden. You manage the infrastructure, backups, monitoring, and scaling.

Best For

Self-hosted deployments needing maximum performance. Replacing Redis clusters with a single Dragonfly instance. Cost-conscious teams with ops capability.

Use Case Recommendations

Caching (Most Common)

Upstash for serverless apps and edge. Redis Cloud for lowest latency. Dragonfly for self-hosted cost savings.

All three work well for caching. Choose based on your deployment model, not caching performance.

Session Storage

Upstash if your app runs on serverless/edge. Redis Cloud for traditional deployments. Both handle sessions well.

Dragonfly works but requires managing persistence and availability yourself.

Rate Limiting

Upstash has a dedicated @upstash/ratelimit SDK designed for serverless rate limiting. Best developer experience for this use case.

Redis Cloud and Dragonfly use standard Redis rate limiting patterns (sliding window, token bucket).

Message Queues

Redis Cloud with Redis Streams for full-featured messaging. Upstash with QStash for serverless messaging. Dragonfly supports Streams but you manage availability.

Real-Time Features (Pub/Sub)

Redis Cloud for persistent connections and pub/sub at scale. Dragonfly for self-hosted pub/sub. Upstash has pub/sub but HTTP-based — add-on products like QStash may be better for async patterns.

Pricing Deep Dive

Upstash

• Free: 10K commands/day, 256MB storage
• Pay-as-you-go: $0.2/100K commands
• Pro: $0.4/100K commands (multi-region, higher limits)
• Enterprise: Custom

Monthly cost examples:

• 1M commands/month: ~$2
• 10M commands/month: ~$20
• 100M commands/month: ~$200

Redis Cloud

• Free: 30MB database
• Fixed: From $7/month (250MB)
• Flexible: From ~$88/month (1GB)
• Annual: Custom pricing (discounts)

Dragonfly

• Self-hosted: Free (your infrastructure costs)
• Dragonfly Cloud: Emerging (check current pricing)
• Typical self-host cost: $20-50/month for a VPS with 4-8GB RAM

Decision Matrix

Scenario	Best Choice
Vercel/Next.js project	Upstash
Cloudflare Workers	Upstash (HTTP API)
Traditional Node.js/Python app	Redis Cloud or Dragonfly
High-throughput, self-hosted	Dragonfly
Need Redis Stack modules	Redis Cloud
Side project / low traffic	Upstash (free tier)
Enterprise with compliance needs	Redis Cloud
Replacing a Redis cluster	Dragonfly

FAQ

Is Upstash actually Redis?

Upstash is Redis-compatible but not Redis itself. It uses a custom implementation with disk-backed storage and HTTP API. For 95% of use cases, the difference is invisible. For ultra-low-latency or advanced Redis module needs, it matters.

Can Dragonfly replace a Redis cluster?

Often yes. A single Dragonfly instance on a 16-core machine can match the throughput of a 6-node Redis cluster, at a fraction of the operational complexity and cost.

What about Valkey (the Redis fork)?

Valkey is the Linux Foundation fork of Redis created after the license change. It's a direct Redis fork maintained by AWS, Google, and others. Consider it for self-hosted deployments where you want true Redis open-source (BSD license). Managed Valkey services are emerging on AWS (ElastiCache) and others.

Is the Redis license change a problem?

For most users: no. You can still use Redis freely. The concern is mainly for cloud providers who want to offer Redis-as-a-service. If license purity matters to you, use Valkey or Dragonfly.

The Verdict

• Upstash for serverless, edge, and pay-per-use. The modern choice for modern architectures.
• Redis Cloud for full Redis compatibility, lowest latency, and enterprise requirements.
• Dragonfly for self-hosted performance, replacing Redis clusters, and cost optimization.

For most developers building on Vercel, Cloudflare, or any serverless platform in 2026: Upstash is the default choice. For traditional server-based apps needing sub-millisecond latency: Redis Cloud. For self-hosted cost savings: Dragonfly.