This topic provides answers to some frequently asked questions about job running in MaxCompute.
Why does a job slowly run and how do I resolve this issue?
The issue may be caused by insufficient resources, job exceptions, or mode rollback.
Insufficient resources
For a job that uses subscription computing resources, the job may wait for resource allocation if a large amount of runtime data is available, a large number of resources are requested, or the job has a low priority. As a result, the job is slowly run. We recommend that you view the resource consumption in the MaxCompute console by following the instructions in the "View quota consumption" section in Manage quotas in the new MaxCompute console. If the resource usage reaches the upper limit of available resources for a long period of time, you must optimize job execution or change the configurations of computing resources. For more information about how to optimize job execution, see Manage jobs in the new MaxCompute console.
For a job that uses pay-as-you-go computing resources, take note that the pay-as-you-go resource pool is a shared resource pool. Computing jobs preempt resources in the pool on demand. You cannot specify the amount of resources that are used by computing jobs. If a large number of jobs are run at the same time, the requested resources are insufficient due to multi-user preemption. As a result, slow response issues occur.
Job exceptions
Job exceptions include data skew caused by jobs, low execution efficiency of user-defined functions (UDFs), and data bloat. For SQL jobs, you can identify the causes of issues by using Logview. For more information about how to identify causes of issues, see Diagnose a job that runs slowly. For more information about the computing optimization method, see Optimize SQL statements.
Mode rollback
MaxCompute jobs can be run in MaxCompute Query Acceleration (MCQA) or normal mode. For jobs that need to process a large amount of data and do not need to return query results, only the normal mode can be used. In most cases, if the resources and jobs are normal, the execution durations of the jobs are basically the same.
In most cases, the MCQA mode is used for interactive query jobs that need to process a small amount of data. The job execution speed in MCQA mode is faster than that in normal mode. MaxCompute cannot ensure that a job always runs in MCQA mode. In some cases, an MCQA job may roll back to a normal job. As a result, the job execution speed is slower than the expected speed.