Products · 4 primitives
One deploy.
Four things it can be.
Compute, data, cache and jobs that ship together and run next to your users. Pick one or use all four.
Trace a request
Watch one request move
through the stack.
01 / FUNCTIONS
TypeScript, Rust and Go on V8 isolates and WASM. Every function is deployed to every city.
export default async (req) => {
return Response.json({ city: req.city })
}
Explore Functions
COLD START
< 5 ms
RUNTIMES
TS, Rust, Go, WASM
MAX DURATION
15 min
MEMORY
1 GB
02 / EDGE POSTGRES
A real Postgres primary with read replicas in 40 regions. Branches spin up per pull request.
const cart = await db.query(
“select * from carts where id = $1”, [id]
)
Explore Edge Postgres
REPLICAS
40 regions
REPLICA LAG
p99 120 ms
BRANCHING
Per pull request
STORAGE
Up to 4 TB
03 / SMART CACHE
Cache responses by key, tag or query. Purges reach all 310 cities in 150 ms.
return cache.wrap(“cart:” + id, 30, () =>
loadCart(id)
)
Explore Smart Cache
GLOBAL PURGE
150 ms
HIT RATIO
94% median
KEYS
Tags, wildcards
SIZE
200 GB
04 / QUEUES
Durable jobs that run where their data lives, with retries and dead letters built in.
await queue.send(“receipts”, { orderId })
// retries: 5, backoff: exponential
Explore Queues
DELIVERY
At least once
RETRIES
Configurable
SCHEDULES
Cron and delays
THROUGHPUT
50k jobs / s
Better together
One binding. No glue code.
STEP 1
Function
handles the request →
STEP 2
Cache
answers in 1 ms →
STEP 3
Postgres
reads the nearest replica →
STEP 4
Queue
runs the slow part later →