Content

Target customers: all users. New features/specifications: In ultra-large-scale MaxCompute jobs, some Fuxi Instances may become long-running due to issues such as data skew or infinite loops. This prolongs the overall job execution time and can even lead to timeout failures. By default, MaxCompute does not retain any result data for failed jobs, resulting in a waste of time and computing resources. In this scenario, MaxFrame allows you to submit partial jobs. This allows you to ignore and lose the entire job in a small range. Even if the job fails, you can obtain the result data of successfully run Fuxi instances.