Beijing MosoInk Information Technology Co., Ltd. is an innovative company that focuses on technology research and platform operations in digital publishing and mobile education. Relying on its self-developed core technology system, MosoInk provides mobile interactive digital teaching materials for publishers, mobile learning platforms for schools, and self-service digital publishing for teachers.

1. The resource capacity was difficult to plan, leading to resource waste.
One of the most difficult problems faced by MosoInk was related to video processing. MosoInk had to process a large number of course videos every day and perform complex editing tasks such as trimming, splitting, merging, transcoding, resolution adjustment, and client adaptation. Over the past several years, MosoInk has developed a controllable video processing mechanism by using technologies such as FFmpeg to support rapid business development. However, the business growth rate this year went beyond the expectation of the MosoInk team. The number of video processing requests during peak hours was dozens of times as many as that in previous years. Overwhelmed by the request volume, the architecture developed by MosoInk provided poor user experience.

2. Horizontal scaling was neither convenient nor fast.
The first solution that the MosoInk technical team came up with was scaling. Horizontal scaling can solve the problem in the short term. However, the number of requests for video processing can fluctuate wildly throughout the day. Most requests must be processed during peak hours. A lot of computing resources would be wasted if the system capacity is planned based on the peak demand. Besides, the peak demand is difficult to predict. If the system schedules resources to process some peak-hour requests during off-peak hours, the cost of computing resources is reduced. However, this may result in long processing queues for customers. The horizontal scaling solution was therefore not adopted.

The team went on to consider improving video processing efficiency by using solutions such as the Backend-as-a-Service (BaaS) solution to receive video transcoding services from cloud providers. These solutions eliminated the need to plan the resource capacity. Instead, MosoInk can consume cloud services based on the actual business requirements. This solution seemed feasible, but MosoInk was hesitant about whether to take it. The video processing mechanism adopted by MosoInk was highly sophisticated. It contained a lot of custom business logic that cannot be implemented by third-party video transcoding services.

MosoInk needed a solution that was cost-efficient, flexible, and O&M-free. These core demands can be satisfied by a serverless service. After extensive research on serverless services provided by Chinese cloud service providers, the MosoInk technical team unanimously decided that the video processing solutions of Alibaba Cloud Function Compute best suited their needs.

Function Compute is an event-driven, fully managed serverless computing service that allows users to write and upload code without having to manage infrastructure resources such as servers. Function Compute prepares computing resources and runs tasks elastically and reliably. With support for a variety of major programming languages, Function Compute allows the MosoInk technical team to seamless migrate the code logic of their original architecture to Function Compute at a low cost.

Function Compute automatically triggers instances to process videos that are uploaded to Alibaba Cloud Object Storage Service (OSS). After Serverless Workflow is integrated, Function Compute can orchestrate distributed tasks. This allows complex operations on large files, such as splitting a file, processing the small files in parallel, and merging the small files. This further improves the processing efficiency.

Function Compute provides a completely on-demand serverless solution, which ensures that the elasticity of cloud computing can be fully utilized. Function Compute can rapidly mobilize the computing resources of tens of thousands of instances to process videos. It also greatly reduces operations costs and eliminates the need to reserve computing resources or maintain the underlying software and hardware. This frees up the MosoInk technical team to focus on the implementation of complex business logic. Compared with traditional solutions, the serverless solution based on Function Compute helps MosoInk save IT service costs by about 60% in video processing scenarios.

The MosoInk technical team wants to integrate Function Compute into even more scenarios to improve the competitiveness of the entire team and better utilize the benefits of cloud native computing.