# Large group chat

* [Differences between livecast room vs 10K group chat](#differences-between-livecast-room-vs-10k-group-chat)
  * [Number of participants](#number-of-participants)
  * [Relationship between user and group](#relationship-between-user-and-group)
  * [Duration of groups](#duration-of-groups)
* [Challenges of livecast room](#challenges-of-livecast-room)
* [Bottleneck of large group chat](#bottleneck-of-large-group-chat)
  * [Batch requests to get presence status](#batch-requests-to-get-presence-status)
  * [Batch requests to the same client gateway](#batch-requests-to-the-same-client-gateway)
  * [Handle online users separately from offline users](#handle-online-users-separately-from-offline-users)
  * [Save offline messages asynchrnously](#save-offline-messages-asynchrnously)
  * [Different queue speed for different group](#different-queue-speed-for-different-group)
* [References](#references)

## Differences between livecast room vs 10K group chat

### Number of participants

* For pure 10K group chat scenarios, 10K is already super big group
* For livecast room scenarios, 10K is pretty common, and it could be as high as million or 10 millions of participates.
  * 1M or 10M participates

### Relationship between user and group

* For pure 10K group chat scenarios, the frequency of joining/leaving groups are pretty low.
* For livecast room scenarios, the frequency of joining/leaving livecast rooms are pretty high.
  * 10K/s-20K/s joining/leaving livecast room per second.

### Duration of groups

* For pure 10K group chat scenarios, group memberships could last at least for months or years.
* For livecast room scenarios, group memberships could only last a few hours.

## Challenges of livecast room

* Latency: Livecast room requires realtime interactions and low latency in APIs.
* End user experience: From end user perspective, each screen could fit 10-20 messages. If there are more than 20 messages per second pushed down to user device, the screen will stuck in a refreshing loop, resulting in bad user experience.

## Bottleneck of large group chat

### Batch requests to get presence status

### Batch requests to the same client gateway

### Handle online users separately from offline users

### Save offline messages asynchrnously

![](https://1010073591-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F-Mk8dv8Mfudl_6ziUzDf%2Fuploads%2Fgit-blob-8f66c61ac825db0daff8aa98ac08e913d4077dec%2Fim_connectionlayer_scale.png?alt=media)

### Different queue speed for different group

![](https://1010073591-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F-Mk8dv8Mfudl_6ziUzDf%2Fuploads%2Fgit-blob-a3d7cf3c732035a5db8d51de9dfc0b34fa07b597%2Fim_largegroup_messagequeue.png?alt=media)

## References

* [大规模群消息推送如何保证实时性?](https://mp.weixin.qq.com/s?__biz=MzI1ODY0NjAwMA==\&mid=2247483830\&idx=1\&sn=b049558916512f4e1d4f34c5489a1d46\&chksm=ea044b29dd73c23fae4508d555446a57c517dcc0c0b1ad47ad0003bb8df44cf04b8d6f1be3ad\&scene=21#wechat_redirect)
* [IM技术分享：万人群聊消息投递方案的思考和实践](https://cloud.tencent.com/developer/article/1869748)