Starting point literary network case

Last updated: 2020-02-26 20:03:25

PDF

The Qidian.com under China Reading Limited meets its 3 core demands through CMQ:

  1. The operating system of "righteous book wealth", in which the function of grabbing monthly tickets in red envelopes is asynchronous when consumers enter the account. The entry information will be written into MQ first. After the consumer pulls, and the consumer confirms that the consumption has been successful, the callback API deletes the information in the MQ.

  2. The journal logs of systems of Qidian.com, including the Olympus M, alarm, and operation systems, will be gathered in the CMQ first, and the big data analysis cluster in the backend will continuously pull messages from CMQ and analyze them based on the processing capacity.

  3. Rewind's ability to provide messages similar to kafka. When the business consumes successfully and deletes the message, you can use the message rewind to re-consume the deleted message. You can specify the location of the offset to adjust. This is convenient for starting point literature network, do bill reconciliation, business system retry and so on.

With a QPS for API requests of over 100000 and a request volume per day of over 1 billion, the pressure imposed on CMQ by the oPCll business of Qidian can be huge. Customers may wonder if CMQ can provide stable support for such a huge business volume?

The cluster at the CMQ backend is transparent to users. Imperceptible's, CMQ controller server can schedule and relocate the queue in real time according to the load of the cluster. If the request volume of a queue exceeds the service threshold, controller server of the current cluster, queue routes can be distributed to multiple clusters to improve concurrency, which can theoretically achieve unlimited Message Retention and ultra-high QPS.

Please see the following figure:
! [ 152355] mc.qcloudimg.com/static/img/026d7738b234171b285a98f0e751038a/image.png)