Today, I was going through videos at VMworld 2010. After which I made few points out, which is something interesting to be aware of.
Changes made in vMotion
1. Overhead associated with vMotion is now carried by (VMM) monitor in ASYNC way, till 4.0 it was synchronous (getting pages from monitor to kernel and again sending it to network module)
2. Now from vSphere 4.1 onwards we can perform,
4 concurrent vMotions over 1GBps network
8 concurrent vMotions over 10GBps network
3. Faster maintenance mode evacuations
A host running with 15 ODD virtual machines now takes nearly 80 – 90 Secs to evacuate virtual machines when entering maintenance mode, this is what is said.
Changes in Resource Management
Major focus was in vSphere 4.1 memory management. Whole idea here is to control memory overheads. And this is achieved by memory compression techniques.
Memory compression now happens at VM level as well at host level.
Earlier compression was done with 4K swap pages and was loaded in compression cache, now the same is done at 2K swap pages. ESX will not allocate any physical memory specially for compression cache now, all operations are carried at VM memory level.
When referring to break out Sessions, just came through “Storage DRS”. New term in VMware storage arena. But really good stuff. Not sure, when this will be launched.
The aim is to automate the tedious task of managing space and I/O load across a group of storage devices by intelligent initial placement and by using storage vMotion of virtual disks between data stores. This is analogous to how server DRS uses vMotion for CPU and memory load balancing.
http://www.vmworld.com/docs/DOC-4668