mbrodmann opened a new pull request #86:
URL: https://github.com/apache/incubator-crail/pull/86


   This PR is the third of three planned PRs for implementing resource 
elasticity within Crail. More specifically, the goal is to allow to 
automatically and dynamically adapt the number of running datanodes based on 
different possible parameters (e.g. resource consumption, available storage 
capacities, ...). This requires support for starting and terminating running 
datanodes automatically. This functionality will be implemented mostly within 
the Crail namenode. Therefore, the Crail namenode will not only maintain 
information on the current state of the datanodes and the deployment but will 
also start or terminate datanodes as required. For deciding how to adjust the 
number of datanodes  based on the current state observed in the Crail namenode 
so-called _policies_ are used. 
   
   The following changes will be included within the seperate PRs:
   
   1) This first PR implements a new RPC (for both darpc and narpc) that allows 
to specify a datanode (using IP-address and port number) that should be 
shutdown.
   2) The seconds PR will implement the actual mechanism for shutting down a 
datanode.
   
   3) This third PR implements a simple policy for implementing elasticity 
based on the system's storage consumption. It introduces a new 
`ElasticNameNodeService` extending the existing `NameNodeService` which 
frequently evaluates the current storage utilization. When it exceeds a certain 
threshold this will make the namenode start new datanode(s). In case the 
utilization falls below a certain threshold the namenode tries to terminate 
unneeded datanodes using the shutdown mechanism. Parameters are configurable.


----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

For queries about this service, please contact Infrastructure at:
us...@infra.apache.org


Reply via email to