Framework for parallel IPL (boot) and error aggregation acoss nodes in a multi-node server
Publication Date: 2016-Jan-06
The IP.com Prior Art Database
Described is a Framework for parallel IPL (boot) and error aggregation acoss nodes in a multi-node server
Page 01 of 1
Framework for parallel IPL (boot) and error aggregation acoss nodes in a multi -node server
In a multi-node server (each node is a separate enclosure with its own hardware), each node needs to go through an Initial Program Load (IPL) in order to boot up. Parts of the IPL can be run in parallel across nodes to minimize boot time. Different components of the server firmware will own different functions of the IPL. There is a need for a common framework via which these components cant get their functions to run in parallel across nodes. Different types of errors might be encountered across nodes for the same IPL function. There is also a need to be aggregate such errors and to correlate them to the IPL function.
Such a framework enables to write firmware for performing IPL in parallel across Nodes in a multi-node server as well as to aggregate errors across Nodes. The Framework takes care of dispatching an IPL function across nodes in parallel, waiting for the functions to finish, and aggregating errors reported by the functions.
The following describes the flow of an IPL function that needs to run in parallel across nodes :
- User registers their IPL function (as a callback) via an interface provided by the framework. The user's IPL function should accept a handle to a node in the server as input.
- User, via an interface, asks the framework to execute the IPL function.
- Based on the number of functional nodes, framework creates worker threads.
- For each functio...