For Redis / Valkey

Redis on Edge,
no per-hour cache fee

Self-hosted Redis or Valkey on real RAM. AOF + RDB persistence to S3-compatible storage, replicas over private networking. The cache, queue and key-value store you already use — yours.

Start your trial Read the deploy guide

# Provision a Redis VM

$ edge compute create \

--image ubuntu-24-04 --plan small \

--script ./bootstrap-redis.sh

# redis.conf — persistence + bind

bind 10.0.0.0/8

appendonly yes

save 900 1

save 300 10

# Snapshot → Edge Storage (cron)

$ aws s3 cp /var/lib/redis/dump.rdb \

s3://redis-snapshots/$(date +%F).rdb \

--endpoint storage.edge.run

Why Redis users move to Edge

Sub-millisecond cache, dedicated RAM, none of the surprise bills.

Real RAM, real CPU

A dedicated VM means Redis isn't fighting for cache or memory bandwidth with noisy neighbours. Sub-millisecond p99s without the Elasticache fees.

AOF + RDB to S3

Run AOF for durability and snapshot RDB to Edge Storage on a schedule. Disaster recovery without paying for a hosted Redis just for the backup feature.

Replicas + Sentinel

Spin up replicas on separate Edge VMs over private networking, manage failover with Sentinel. Standard Redis HA, no proprietary plumbing.

Cluster mode when you need it

Redis Cluster across multiple VMs gives you sharding for huge keyspaces. Or stick with a single primary — most workloads never outgrow it.

Redis or Valkey

Same install pattern, swap the binary. Use Valkey if the licensing change matters to you, vanilla Redis otherwise. Both are first-class on Edge VMs.

No per-hour cache markup

Elasticache cache.r6g.large is ~$130/month. The same RAM on an Edge VM is a fraction of that — and you keep the keys.

Reference architecture

How Redis maps to Edge

Primary on a small VM with plenty of RAM, optional replicas + Sentinel for HA, snapshots to object storage.

Compute

Redis (or Valkey) on a VM, optional Sentinel for HA

Storage

RDB snapshots and AOF rewrites archived to a bucket

DNS

Anycast DNS for `cache.example.com` pointing at the primary

Compute (replica)

Optional read replicas on private networking

# From a Laravel / Rails / Node app

REDIS_URL=redis://cache.private.edge.network:6379

# Standard client libs work unchanged

await redis.set('user:42', JSON.stringify(user))

const cached = await redis.get('user:42')

# Sub-ms over the private network

PING → PONG (0.3ms)

Common questions

How does this compare to Elasticache / Memorystore?

Cheaper (no per-hour cache fees), faster on like-for-like hardware (real CPU, real RAM), and you keep root access. Trade-off: you handle backups and Sentinel — or our Expert Services team can.

Should I use Redis or Valkey?

Functionally identical for almost all workloads. Use Valkey if the BSL/SSPL licence change in Redis 7.4+ is a concern; use Redis otherwise. Both are drop-in compatible.

How do I do persistence?

Combine AOF (every-write durability) with periodic RDB snapshots to Edge Storage. For pure-cache use cases, you can skip persistence entirely and treat the VM as ephemeral.

How do I scale beyond one VM?

Read scaling: spin up replicas on more VMs. Write scaling: Redis Cluster across multiple primaries. Both work over Edge's private networking with sub-millisecond latency.

By Stack