July 18, 2024
Actual-time Messaging – Slack Engineering

Do you know that floor stations transmit indicators to satellites 22,236 miles above the equator in geostationary orbits, and that these indicators are then beamed all the way down to your complete North American subcontinent? Satellite tv for pc radios as we speak serve a whole bunch of channels throughout 9,540,000 sq. miles. Except you’re working at a secret army facility, deep underground, you may take pleasure in satellite tv for pc radio all over the place. 

Identical to the satellites, Slack sends tens of millions of messages daily throughout tens of millions of channels in actual time all around the globe. If we take a look at the visitors on a typical work day, it reveals that the majority customers are on-line between 9am and 5pm native time, with peaks at 11am and 2pm and a small dip in between for lunch hour. Although the working hours are related throughout areas, trying on the two peaks within the graph beneath, it’s evident that prime time will not be the identical: It’s post-noon in some areas and pre-noon in different areas. Every coloured line within the beneath graph represents a area.



On this weblog submit we’ll describe the structure that we use to ship real-time messages at this scale. We’ll take a more in-depth take a look at the providers that ship the chat messages and varied occasions to those on-line customers in actual time. Our core providers are written in Java: They’re Channel Servers, Gateway Servers, Admin Servers, and Presence Servers.

Server overview

Channel Servers (CS) are stateful and in-memory, holding some quantity of historical past of channels. Each CS is mapped to a subset of channels primarily based on constant hashing. At peak instances, about 16 million channels are served per host. A “channel” on this occasion is an summary time period whose ID is assigned to an entity similar to person, crew, enterprise, file, huddle, or a daily Slack channel. The ID of the channel is hashed and mapped to a singular server. Each CS host receives and sends messages for these mapped channels. A single Slack crew has all of its channels mapped throughout all of the CSs.

Constant hash ring managers (CHARMs) handle the constant hash ring for CSs. They substitute unhealthy CSs in a short time and effectively; a brand new CS is able to serve visitors in underneath 20 seconds. With a crew’s channels unfold throughout all CSs, a small variety of groups’ channels are mapped to a CS. When a channel server is changed, customers of these groups’ channels expertise elevated latency in message supply for lower than 20 seconds.

The diagram beneath reveals how CSs are registered in Consul, our service discovery instrument. Every constant hash is outlined and managed by CHARMs, after which Admin Servers (AS) and CS discovers them by querying Consul for the up-to-date config.



Gateway Servers (GS) are stateful and in-memory. They maintain customers’ info and websocket channel subscriptions. This service is the interface between Slack purchasers and CSs. Not like all different servers, GSs are deployed throughout a number of geographical areas. This permits a Slack consumer to shortly hook up with a GS host in its nearest area. We’ve a draining mechanism for area failures that seamlessly switches the customers in a foul area to the closest good area.

Admin Servers (AS) are stateless and in-memory. They interface between our Webapp backend and CSs. Presence Servers (PS) are in-memory and maintain observe of which customers are on-line. It powers the inexperienced presence dots in Slack purchasers. The customers are hashed to particular person PSs. Slack purchasers make queries to it by means of the websocket utilizing the GS as a proxy for presence standing and presence change notifications. A Slack consumer receives presence notifications just for a subset of customers which can be seen within the app display at any second.

Slack consumer arrange

Each Slack consumer has a persistent websocket connection to Slack’s servers to obtain real-time occasions to keep up its state. The consumer units up a websocket connection as beneath.



On boot up, the consumer fetches the person token and websocket connection setup info from the Webapp backend. Webapp is a Hacklang codebase that hosts all of the APIs referred to as by our Slack Shoppers. This service additionally contains JavaScript code that renders the Slack purchasers. A consumer initiates a websocket connection to the closest edge area. Envoy forwards the request to GS. Envoy is an open supply edge and repair proxy, designed for cloud-native functions. Envoy is used at Slack as a load-balancing resolution for varied providers and TLS termination. GS fetches the person info, together with all of the person’s channels, from Webapp and sends the primary message to the consumer. GS then subscribes to all of the channel servers that maintain these channels primarily based on constant hashing asynchronously. The Slack consumer is now able to ship and obtain actual time messages.

Ship a message to one million purchasers in actual time

As soon as the consumer is about up, every message despatched in a channel is broadcasted to all purchasers on-line within the channel. Our message stats reveals that the multiplicative issue for message broadcast is totally different throughout areas, with some areas having the next charge than others. This may very well be as a consequence of a number of elements, together with crew sizes in these areas. The chart beneath reveals message acquired depend and message broadcasted depend throughout a number of areas.



Let’s check out how the message is broadcasted to all on-line purchasers. As soon as the websocket is about up, as mentioned above, the consumer hits our Webapp API to ship a message. Webapp then sends that message to AS. AS appears to be like on the channel ID on this message, discovers CS by means of a constant hash ring, and routes the message to the suitable CS that hosts the true time messaging for this channel. When CS receives the message for that channel, it sends out the message to each GS the world over that’s subscribed to that channel. Every GS that receives that message sends it to each linked consumer subscribed to that channel id.

Beneath is a journey of a message from the consumer by means of our stack. Within the following instance, Slack consumer A and B are in the identical edge area, and C is in a unique area. Consumer A is sending a message, and consumer B and C are receiving it.



Other than chat messages, there’s one other particular form of message referred to as an occasion. An occasion is any replace a consumer receives in actual time that adjustments the state of the consumer. There are a whole bunch of various kinds of occasions that circulate throughout our servers. Some examples embrace when a person sends a response to a message, a bookmark is added, or a member joins a channel. These occasions comply with an identical journey to the easy chat message proven above. 

Have a look at the message supply graph beneath. The depend spikes at common intervals. What might trigger these spikes? Seems, occasions despatched for reminders, scheduled messages, and calendar occasions are likely to occur on the high of the hour, explaining the common visitors spikes.



Now let’s check out a unique form of occasion referred to as Transient occasions. These are a class of occasions that aren’t continued within the database and are despatched by means of a barely totally different circulate. Consumer typing in a channel or a doc is one such occasion.



Beneath is a diagram that reveals this situation. Once more, Slack consumer A and B are in the identical edge area, and C is in a unique area. Slack consumer A is typing in a channel and that is notified to different customers B and C within the channel. Consumer A sends this message through websocket to GS. GS appears to be like on the channel ID within the message and routes to the suitable CS primarily based on a constant hash ring. CS then sends to all GSs the world over subscribed to this channel. Every GS, on receiving this message, broadcasts to all of the customers websockets subscribed to this channel


What’s subsequent

Our servers serve tens of tens of millions of channels per host, tens of tens of millions of linked purchasers, and our system delivers messages the world over in 500ms. With the linear scalability of our present structure, our projections present that we are able to serve many extra clients. Nonetheless, there’s all the time room for enchancment and we need to prolong our structure to serve the dimensions of our subsequent largest clients. If this work sounds attention-grabbing to you, come be a part of us: we have now an open role !

Lastly, an enormous shout out to everybody who contributed to this structure, and to Serguei Mourachov for reviewing and giving suggestions on this weblog submit.