Method and apparatus for resource scheduling simulation stress

Abstract

The purpose of this application is to provide a method and device for resource scheduler to simulate pressure, obtain the resource request sequence submitted by the resource request sequence generator through the resource scheduler, and then search for a resource node and request to schedule the resource node The user resource manager parses the resource request sequence; the user resource manager parses the resource request sequence into several user resource request units and sends several user resource request units and corresponding resources to the resource scheduler; resource The scheduler allocates the received user resource request units and corresponding resources to one or more resource nodes for simulated pressure processing; finally, the user resource manager calculates the resources used by the simulated user according to the resources allocated by the resource scheduler. The user resource requests the time required by the unit to perform calculations. The resource scheduler and the user resource manager run on different physical nodes, so as to realize the scalability of the pressure simulator.

Description

technical field [0001] The present application relates to the field of computers, and in particular to a technology for resource scheduling simulation pressure. Background technique [0002] In a distributed system, the resource scheduler not only needs to manage tens of thousands of resource nodes running on physical nodes in the cluster, but also needs to undertake tens of thousands of resource scheduling requests per second, and needs to do so at the time complexity of milliseconds make a decision. So it is very important to observe and evaluate the behavior characteristics of the resource scheduler under pressure. [0003] Since the current industrial-level cluster size has reached more than 5,000, it is not only difficult to run a resource scheduler under such a scale of physical nodes, but also very high in terms of maintenance costs. Therefore, a technical means that can simulate real large-scale resource nodes on small-scale physical nodes is needed. [0004] Curr...

AI Technical Summary

Problems solved by technology

In the technical implementation of this solution, the scheduler and the simulator run on the same physical machine, making it impossible to achieve distribution, and it is also impossible to expand the pressure of the simulator to multiple physical nodes, so the real cluster size that can be simulated is limited. In addition, SLS can only receive pressure sequence data in a fixed format, so that data from different sources need to be pre-processed before being input to SLS. For example, the log data of online clusters and artificially generated data formats are often different. thus limiting its flexibility

Images