Surety is performing system maintenance this weekend. Electronic date stamps on new Prior Art Database disclosures may be delayed.
Browse Prior Art Database

Enhancement for the SOA Workload Scheduler

IP.com Disclosure Number: IPCOM000241411D
Publication Date: 2015-Apr-24
Document File: 2 page(s) / 55K

Publishing Venue

The IP.com Prior Art Database


Disclosed is a method to increase efficiency in a Service Oriented Architecture by running failed tasks on alternate hosts and adding a new task retry limit counter, which is only for the host-unreachable case.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 58% of the total text.

Page 01 of 2

Enhancement for the SOA Workload Scheduler

For a Service Oriented Architecture (SOA) workload with millions of tasks, the user (i.e. customer) usually builds a cluster with hundreds of hosts and runs tasks on these hosts. For an error handling case, if a task fails (e.g., throws an exception, returns unexpected results, etc.), the system runs the task again to produce the expected result. If a host becomes unreachable (e.g., network glitch), then the tasks on this host are run again on another host.

This process has two problems:

Issue 1: Some tasks only succeed on some hosts, while failing on other hosts (e.g., cannot access data on some hosts); therefore, if a failed task is run again on the same host, then the task fails again, which wastes the central processing unit(CPU) time.

Issue 2: Some tasks have the wrong program logic, which crashes the host on which said task are running; thus, this host becomes unreachable. If such tasks are run again on other hosts, then all the hosts in the cluster crash in sequence and become unreachable. This prevents the cluster from running the workload.

The novel solution to Issue 1 is to run the failed tasks on another host, providing such tasks the opportunity to complete. The hosts with failed tasks can still be used to run other tasks; the resource is not wasted.

The novel solution to Issue 2 is to add a new task retry limit counter, which is only for the host-unreachable case. A re-run task's retry limit counter can be...