From 9bd2a260815a2441332285d32ccfcbfa466e8f3e Mon Sep 17 00:00:00 2001 From: Helin Wang Date: Fri, 10 Nov 2017 14:10:46 -0800 Subject: [PATCH] Rename Design Doc: Horizontal Autoscaling to Design Doc: Elastic Deep Learning --- doc/autoscale/README.md | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) diff --git a/doc/autoscale/README.md b/doc/autoscale/README.md index 40440c92..09ad22fc 100644 --- a/doc/autoscale/README.md +++ b/doc/autoscale/README.md @@ -1,4 +1,4 @@ -# Design Doc: Horizontal Autoscaling +# Design Doc: Elastic Deep Learning ## Background @@ -6,11 +6,11 @@ A PaddlePaddle training job contains several trainer instances, several parameter server instances, and one master instance. We would like to automatically scale the number of training instances and the number of parameter server instances to fully utilize the cluster's -computation resources. +computation resources. We call this Elastic Deep Learning. -Currently, we will only support horizontal trainer -autoscaling. Parameter server horizontal autoscaling will be supported -in the near future. This design doc considers both of them. +Currently, we will only support trainer autoscaling. Parameter server +autoscaling will be supported in the near future. This design doc +considers both of them. [Horizontal Pod Autoscaling (HPA)](https://kubernetes.io/docs/tasks/run-application/horizontal-pod-autoscale/) is an autoscaling solution provided by Kuberentes, but it's not suitable