Author: Busy fish technology - Jin Yi
background
Busy fish as an electricity supplier scene app, the richest part is browsing feeds hosted as a commodity baby, such as home following baby flow of information, the search results page and the details page below you may also like these feeds scenes and ultimately recommendation algorithm in support behind it.
The traditional recommendation algorithm is relying on the cloud point data buried precipitation to. With the decline 4G network tariff, resulting in more and more data, if such a large amount of data uploaded to the operation, and centralized storage / computing and processed by the cloud to do, not only will generate a lot of unnecessary network flow, but also bring high costs to cloud storage, real-time can not be guaranteed.
We observed that with the upgrading of mobile phone computing power, certain calculations can be directly calculated at the end, and then reported to the unified back-end, so the cloud has many obvious advantages compared to:
1, more timely: We can complete the original feature at the end of processing and real-time scoring, extracted from the original features of the reported results is completed, it can be completed in about 1s.
2, computing resource conservation: If large amounts of data are gathered to the cloud meter