Nov 16, 20268 min read

Upstash Alternatives in 2026: Managed Redis for Backends in Your Cloud

The main alternatives to Upstash Redis, how they compare on pricing model, networking, consistency and limits, and how to declare a cache in code that runs on ElastiCache or Memorystore.

Illustration for Upstash Alternatives in 2026: Managed Redis for Backends in Your Cloud

Upstash made Redis work from serverless and edge functions by putting an HTTP API in front of it and billing per command, so a database costs nothing when idle. That model starts to fit worse when traffic is steady, when the backend runs on long-lived containers in AWS or GCP and would rather reach a cache on its private network, or when a workload reaches the limits of a shared plan.

This guide compares the main Upstash alternatives on pricing model, networking, consistency and limits, and shows how Encore declares a cache in your Go code and runs it on ElastiCache or Memorystore in your own cloud account.

Upstash alternatives at a glance

Runs inPricing modelReached overBest for
EncoreElastiCache or Memorystore in your AWS or GCP accountYour cloud billPrivate network, Redis protocolBackends on AWS or GCP with the cache declared in code
Amazon ElastiCacheYour AWS VPCPer node-hour, or serverless per GB-hour and ECPUPrivate network, Redis protocolBackends already on AWS
Google MemorystoreYour GCP VPCPer instance capacityPrivate IP, Redis protocolBackends on Cloud Run or GKE
Redis CloudRedis Inc.'s infrastructure on AWS, GCP or AzurePer plan and capacityInternet with TLS, or peeringRedis features and multi-region across clouds
Cloudflare Workers KVCloudflare's networkPer read, write and GB storedWorkers binding or HTTPRead-heavy data at the edge
UpstashUpstash on AWS, GCP or FlyPer command, or fixed plans from $10/monthHTTP REST API or Redis protocol over TLSServerless and edge functions

Upstash prices are from its Redis pricing page, checked November 16, 2026: the Free plan includes 250 MB and 500,000 commands a month, pay as you go bills $0.20 per 100,000 commands with the first 1 GB of storage free and $0.25 per GB after that, fixed plans start at $10 a month for 250 MB, and the Prod Pack with an uptime SLA and Multi-Zone HA adds $200 a month per database.

1. Encore

Encore

Encore is a backend framework with built-in infrastructure for Go and TypeScript, and in Go a cache is a cluster and typed keyspaces declared next to the code that uses them. encore run starts an in-memory Redis locally, and in AWS and GCP environments Encore provisions the cache as Amazon ElastiCache or Google Memorystore in your own account, with the connection wired into your services and nothing to configure per environment.

Set up the app

Get started
brew install encoredev/tap/encore &&
encore app create
Copy

Run in your terminal to get started locally.

~
$ encore app create geoapi --example=hello-world$ cd geoapi && mkdir geo

A common Upstash workload is a per-key rate limit plus a cache in front of a slow or metered API, so the geo service does both:

// geo/geo.go package geo import ( "context" "errors" "strings" "time" "encore.dev/beta/errs" "encore.dev/storage/cache" ) // ElastiCache on AWS, Memorystore on GCP, an in-memory Redis with encore run. var cluster = cache.NewCluster("geo", cache.ClusterConfig{ EvictionPolicy: cache.AllKeysLRU, }) type windowKey struct { APIKey string Minute int64 } // Fixed-window rate limit: one counter per API key per minute. var requests = cache.NewIntKeyspace[windowKey](cluster, cache.KeyspaceConfig{ KeyPattern: "ratelimit/:APIKey/:Minute", DefaultExpiry: cache.ExpireIn(2 * time.Minute), }) type Place struct { Query string `json:"query"` Lat float64 `json:"lat"` Lng float64 `json:"lng"` } // Cached lookups, so repeated queries skip the slow upstream call. var places = cache.NewStructKeyspace[string, Place](cluster, cache.KeyspaceConfig{ KeyPattern: "place/:key", DefaultExpiry: cache.ExpireIn(24 * time.Hour), }) const limitPerMinute = 60 type LookupParams struct { APIKey string `header:"X-API-Key"` Query string `query:"q"` } type LookupResponse struct { Place Place `json:"place"` Cached bool `json:"cached"` } //encore:api public method=GET path=/geo/lookup func Lookup(ctx context.Context, p *LookupParams) (*LookupResponse, error) { if p.APIKey == "" { return nil, &errs.Error{Code: errs.Unauthenticated, Message: "missing X-API-Key"} } q := strings.ToLower(strings.TrimSpace(p.Query)) if q == "" { return nil, &errs.Error{Code: errs.InvalidArgument, Message: "q is required"} } n, err := requests.Increment(ctx, windowKey{p.APIKey, time.Now().Unix() / 60}, 1) if err != nil { return nil, err } if n > limitPerMinute { return nil, &errs.Error{Code: errs.ResourceExhausted, Message: "rate limit exceeded, retry next minute"} } if hit, err := places.Get(ctx, q); err == nil { return &LookupResponse{Place: hit, Cached: true}, nil } else if !errors.Is(err, cache.Miss) { return nil, err } place, err := geocode(ctx, q) if err != nil { return nil, err } if err := places.Set(ctx, q, place); err != nil { return nil, err } return &LookupResponse{Place: place, Cached: false}, nil } // geocode stands in for a slow or metered upstream API. func geocode(ctx context.Context, q string) (Place, error) { time.Sleep(200 * time.Millisecond) return Place{Query: q, Lat: 48.2082, Lng: 16.3738}, nil }
~/geoapi
$ encore run$ curl -H 'X-API-Key: k1' 'localhost:4000/geo/lookup?q=Vienna'{"place": {"query": "vienna", "lat": 48.2082, "lng": 16.3738}, "cached": false}$ curl -H 'X-API-Key: k1' 'localhost:4000/geo/lookup?q=Vienna'{"place": {"query": "vienna", "lat": 48.2082, "lng": 16.3738}, "cached": true}

The second call comes from the cache, a request without a key gets a 401, and the 61st request from one key within a minute gets a 429 with resource_exhausted. Increment is atomic in Redis, so the limit holds across every instance of the service, and each counter expires two minutes after its window so old windows clean themselves up. The keyspaces are typed: the key pattern is checked against the key struct, and cache.Miss tells a miss apart from a Redis error.

Upstash users often reach for QStash as well, for delayed jobs and scheduled HTTP calls. In Encore those are Pub/Sub subscriptions with retries and cron jobs declared in the same Go code, so the cache, the queue and the schedule are provisioned together.

Best fit: Go backends on AWS or GCP that want the cache on a private network next to their services, declared in code and provisioned per environment.

2. Amazon ElastiCache

ElastiCache runs Valkey, Redis OSS or Memcached inside your VPC, either as node-based clusters where you pick instance types, shards and replicas, or as ElastiCache Serverless, which scales capacity on its own. According to the ElastiCache pricing page, Serverless bills for data stored in GB-hours and for requests in ElastiCache Processing Units, where a read or write uses 1 ECPU per KB transferred, and Serverless for Valkey has 33% lower pricing and a 100 MB minimum per cache against 1 GB for Redis OSS.

Clients connect with standard Redis clients over the VPC, which keeps latency low for code running on ECS, EKS or Lambda in the same VPC, but means the cache cannot be reached from outside without extra networking, and there is no HTTP API for edge functions.

Best fit: backends on AWS with steady traffic, where a private cache next to the compute is cheaper and faster than per-command billing over the internet.

3. Google Memorystore

Memorystore offers managed Valkey, Redis and Redis Cluster on GCP, reached over private IP from Cloud Run, GKE or Compute Engine. Instances can replicate across zones with automatic failover, and pricing follows provisioned capacity rather than commands.

Cloud Run reaches a Memorystore instance through Direct VPC egress or a Serverless VPC Access connector, which is one more piece of setup than an Upstash URL but keeps cache traffic off the public internet.

Best fit: backends on Cloud Run or GKE that want a cache in the same Google Cloud project and region.

4. Redis Cloud

Redis Cloud is the managed service from Redis Inc. on AWS, GCP or Azure, with the company's search, JSON and time series features and active-active replication across regions. It runs on Redis Inc.'s infrastructure, reached over the internet with TLS or through VPC peering and private links on higher plans.

Redis moved away from the BSD license in 2024, which led to the Valkey fork under the Linux Foundation, and Redis Cloud is the main place to get Redis Inc.'s own features as a service, while AWS and Google put Valkey first.

Best fit: teams that need Redis Inc.'s modules or active-active replication across clouds.

5. Cloudflare Workers KV

Workers KV is a key-value store on Cloudflare's network rather than Redis, read from Workers through a binding. According to Cloudflare's KV documentation, reads are cached at the edge and eventually consistent, and a write can take 60 seconds or more to be visible in other locations. The Workers paid plan includes 10 million reads and 1 million writes a month, then $0.50 per million reads and $5.00 per million writes, from Cloudflare's KV pricing page checked November 16, 2026.

That makes it fast and cheap for configuration, feature flags and other data read far more often than written, and a poor fit for counters, rate limits or locks, which need atomic operations that KV does not offer.

Best fit: read-heavy data served from Cloudflare Workers, where eventual consistency is acceptable.

How to choose

Stay on Upstash when the code runs in serverless or edge functions that need Redis over HTTP and traffic is low or spiky enough that per-command pricing wins. Pick ElastiCache or Memorystore when the backend runs on AWS or GCP and steady traffic makes a private, capacity-priced cache cheaper and faster. Pick Redis Cloud for Redis Inc.'s features across clouds, and Workers KV for edge reads that can be eventually consistent. Pick Encore when you want ElastiCache or Memorystore with the cache, its keyspaces and the queues and schedules around it declared in your Go code and running locally with encore run.

Frequently asked questions

What is the best alternative to Upstash?

For a backend that runs on AWS or GCP, a common alternative is a managed Redis or Valkey inside your own VPC: Amazon ElastiCache on AWS or Google Memorystore on GCP. Redis Cloud suits teams that want Redis Inc.'s features across clouds, and Cloudflare Workers KV suits read-heavy data served from the edge.

Why move off Upstash?

Per-command pricing grows with steady traffic, a private cache next to compute keeps Redis traffic off the public internet, and Upstash's [billing docs](https://upstash.com/docs/redis/overall/billing) cap Free, pay-as-you-go and fixed plans up to 50 GB at 10,000 commands per second, or 16,000 on the 100 GB and 500 GB plans. Upstash stays a good fit for serverless and edge functions that need Redis over HTTP.

Is ElastiCache cheaper than Upstash?

It depends on traffic. Upstash pay as you go bills $0.20 per 100,000 commands, so cost grows with every request, while a node-based ElastiCache cluster costs the same per hour at any request rate and ElastiCache Serverless bills for data stored and ECPUs used. Low or spiky traffic often costs less on Upstash, and steady high traffic often costs less on ElastiCache.

Does Encore support Redis?

Yes. In Go you declare a cache cluster and typed keyspaces in code. encore run starts an in-memory Redis locally, and Encore provisions Amazon ElastiCache on AWS or Google Memorystore on GCP for each environment and connects your services to it.

Deploy to your cloud on AWS and Google Cloud
without managing infrastructure

Encore lets developers and agents define application infrastructure directly in code, then automatically provisions it from local development to production in your AWS or GCP account.

~/orders — claude
Claude Code
Claude Code v2.1.180Opus 5.5~/orders