Merge pull request #8 from Azure/docs/autoscale

Add a brief section on autoscale and scalesets

README.md

@@ -1,3 +1,19 @@
+# Azure CycleCloud PBS Professional project
+
+PBS Professional is a highly configurable open source workload manager. See the
+[PBSPro project site](http://www.pbspro.org/) for an overview and the [PBSpro
+documentation](https://www.pbsworks.com/PBSProductGT.aspx?n=Altair-PBS-Professional&c=Overview-and-Capabilities&d=Altair-PBS-Professional,-Documentation)
+for more information on using, configuring, and troubleshooting PBS Professional
+in general.
+
+Azure CycleCloud uses the open source community edition of PBS Professional.
+
+## Autoscale and scalesets
+
+In order to try and ensure that the correct VMs are provisioned for different types of jobs, CycleCloud treats autoscale of MPI and serial jobs differently in PBS Professional clusters. 
+
+For serial jobs, multiple VM scalesets (VMSS) are used in order to scale as quickly as possible. For MPI jobs to use the InfiniBand fabric for those instances that support it, all of the nodes allocated to the job have to be deployed in the same VMSS. Currently, a single VMSS is used for all MPI jobs. This can occasionally lead to slower provisioning and deprovisioning of nodes since VMSS operations are atomic. If the scaleset is waiting on some nodes to deprovision, CycleCloud must wait for that operation to complete to provision more nodes.
+
 
 # Contributing