Andrei's answer is pretty comprehensive, and to add a few more things: 1. WRT reliability, YMMV depending on which cloud and ssh implications. For example, most issues we have are swirling around ssh specifically in Amazon EC2. jclouds already has users who start several hundred JVMs (ex. Neotys), but they do not rely on SSH for configuration. It is likely the case you'd want to run Whirr from within the cloud of choice to further reduce issues.
2. to answer you directly, depending on the cloud, the network calls will likely be the same as starting 10, but multiplied by 100, unless some other scaling or pool design is put into place. Most clouds do not have a way to start more than one machine with a single call. There are places, for ex. firewalls, where we could optimize and use a single call for 1000 machines (if they are in the same network, for ex.). Assume these optimizations are not complete, yet. I hope this helps! -Adrian On Thu, Nov 3, 2011 at 2:30 AM, Edward J. Yoon <[email protected]> wrote: > Hi, > > When user want to deploy Hadoop 1 thousand VMs cluster using Whirr and > JClouds, how many API calls e.g., auth + VM creation + firewall > settings, .., etc can be made? > > -- > Best Regards, Edward J. Yoon > @eddieyoon >
