Websocket support

@jdavis Workers are billed on number of “requests”, not on bandwidth nor CPU time. I don’t think we’d want to add CPU or bandwidth metering that applies only to WebSocket. So we need to turn WebSockets into “requests” somehow…

Wasted effort, probably best to just forget websockets in that case. Maybe push this effort into TCP anything?

I think we can find a reasonable way to count WebSockets as requests for billing purposes.

1 Like

So we’re talking about being able to keep websockets open for an extended period of time? If not, what’s the point?

On another note, when is Cloudflare going to purchase ZeroTier and use it’s kernel-bypass expertise to make ZT run screaming fast? In addition, making ZT run within websocket framing on 443 would also be quite disruptive.

Hello @KentonVarda ,

has there been done any progress internally on the Websockets? I would be curious whether we can expect it soon or rather you found complications and decided to postpone it for time being.

Many thanks for reply!

1 Like

@janusz We have a lot on our plate and unfortunately this is not our top priority, but in the time between other things I have managed to create a prototype implementation allowing a Worker to terminate WebSockets as a client or a server. There are still problems with this prototype:

  • In my prototype, an outgoing WebSocket can only be used in the context of the FetchEvent that initiated it. You can’t store it to a global variable and use it across multiple incoming requests. This is because subrequests are intrinsically tied to the FetchEvent that created them in our implementation’s concurrency model. For some use cases this is fine, but for others it may be disappointing. It’ll take a fair amount of work to change this, but it can be done.
  • There’s still no way to extend the CPU timeout beyond 50ms, as mentioned above. Solving this is more of a product question than an engineering question, so requires some discussion within the organization.
  • I’ve only implemented the bare minimum components of the API. The standard defines a wider API surface that needs to be filled out.
  • I have had to extend the Fetch, Service Workers, and WebSocket APIs in non-standard ways, and would like to run my changes past WHATWG before we commit to anything.

Unfortunately since we have a lot of other high-priority things competing for our time, it’s hard for me to say how long it will take to finish this up. It seems like at least a few months away.

I’d be interested to know which of the above constraints would be problems for your use case. If none of them are, then maybe we can have something working for you relatively sooner.

1 Like

My initial idea would be to treat websocket termination more as a high level Cloudflare service similar to how currently http2 and ssl services are handled and expose just message events to worker functions, were the message events have additional matadata properties to know everything relevant about the connection. This way it could be easy to bill 3 or 4 websocket messages as one request and the concurrency and timout characteristics of workers would play nicely.

1 Like

@KentonVarda

I was wondering if there will there be any updates for WebSocket support any time soon?

Thanks in advance

Hi @nesh,

The status is still the same as in my previous comment. We need to figure out how to allow long-running requests and properly bill for them. This has moved slower than I’d hoped, but it’s still a priority. Sorry, I don’t have a date I can tell you.

Ah I see no worries

Is there any update on this in 2020?

1 Like

Not specifically; this is still something we would like to support but do not currently support.

Congrats on all the recent launches this week. Any updates in terms of how Cloudflare is think about adding (or not) WebSocket support, particularly in light of the new billing options?

4 Likes

Looks like websocket support has been added in companion to Durable Objects. While the documentation for durable objects is live, there is not yet any documentation on the WebSocketPair additions to the fetch spec.

Can you comment on pricing for websockets, and how long and how many websockets are supported?

1 Like


What we know, from the beta :

  • WebSockets are billed per connection (1 connection = 1 Worker Request)
  • It does not matter how much data you send one way or the other through the socket
  • You are limited by CPU (50ms with regular Worker), or billed by CPU (Unbound Worker)
  • If the Worker crash, all sessions are invalid, all clients need to reconnect
  • Periodically (TBD), old clients would probably be disconnected because idle
  • Periodically (TBD), old clients would probably be disconnected because new clients
  • If a client is disconnected, it might need to reconnect (1 connection billed)

If you use Websockets for WebRTC, it is probably fine.
Video/Audio will require asking the user to validate a user gesture in their browser.

If you use Websockets for push notifications, it is probably fine.
Depending on your appetite for GAFA dependency, you might want to switch on Apple / Google Push notification system, but it requires asking the user to validate a user gesture in their browser. Ground truth : 90% of push notifications asked by websites are refused by users. It means you can save / offload 10% of your push notification load to GAFA. Of course you’ll need to encrypt those messages to ensure their confidentiality. Depending on volumes, it might be worth (or not).

If you use Websockets as a replacement of HTTP, the best strategy would be :

const doRequest = async (data) => {
  // If there is no WebSocket connection currently active
  //   Open a Websocket connection 
  // Send data over Websocket
}

That way you create a WebSocket connection only when needed, then reuse it until being disconnected, and only reconnect when needed.
If you are exclusively using this mode (no-push-notifications, no-chat, no-WebRTC), then avoid auto-reconnect. It would be a waste of connections.
If you are mixing this usage with other usages (push-notifications, chat, WebRTC), then you’ll have to auto-reconnect because somehow you will need push from server.

Spare connections when user is idle

const f = async () => {
  const {userState, screenState} = detector;
  if (userState === 'idle' || screenState === 'locked') { 
    // If Websocket connected
    //   Close WebSocket
  }
  else { 
    // If there is no WebSocket connection currently active
    //   Open a Websocket connection 
  }
};

const detector = new IdleDetector();
detector.addEventListener('change', f, { passive:true });

Spare connections when page is hidden (minimized, etc)

let hidden, visibilityChange;
if (typeof document.hidden !== 'undefined') { // Opera 12.10 and Firefox 18 and later support
  hidden = 'hidden';
  visibilityChange = 'visibilitychange';
} else if (typeof document.msHidden !== 'undefined') {
  hidden = 'msHidden';
  visibilityChange = 'msvisibilitychange';
} else if (typeof document.webkitHidden !== 'undefined') {
  hidden = 'webkitHidden';
  visibilityChange = 'webkitvisibilitychange';
}

const f2 = async () => {
  if (document[hidden]) {
    // If Websocket connected
    //   Close WebSocket
  } else {
    // If there is no WebSocket connection currently active
    //   Open a Websocket connection 
  }
}

document.addEventListener(visibilityChange, f2, { passive:true });

Keep in mind that a closed connection means you are out of sync thus you don’t receive real time updates. If you need 100% uptime on realtime updates, just stay connected (functionnality being more important than cost impact).

Last point, if you are receiving lots of messages (ex: gaming : positions of players), you might want to apply backpressure on client, to reduce CPU usage. In this case, using the WebSocketStream API is one option.

It works like pulling in local, reading what you received in the buffer, before actually processing it. As it is a pull strategy, you won’t receive updates in real time, since the action of reading your local data is at your hand.

The current example uses a while true strategy, which is common in gaming rendering loops, but not so much in classic web programming.

// This code sequentialize the events : 
// No more parallelization nor real time
// This is how client backpressure works
// Use it only if high loads of events are received
// but your process function takes time 
// and cannot follow the cadence
// (like game engine rendering)

const wss = new WebSocketStream(WSS_URL);
const {readable, writable} = await wss.connection;
const reader = readable.getReader();
const writer = writable.getWriter();

while (true) {
  const {value, done} = await reader.read();
  if (done) {
    break;
  }
  const result = await renderMyGamingEngine(value);
  await writer.write(result);
}
3 Likes

Update on how WebSocket interacts with Durable Object:

3 Likes

This topic was automatically closed 3 days after the last reply. New replies are no longer allowed.