VMware Distributed Power Mgmt (DPM)

[jingling30]

Great demonstration and logic

[ThePinxu]

4. Scalability issue.

We have to use a SAN and disk array in this case. Any time a server is on/off, the storage fabric will see a member join/leave and will do a lot of work. Although the VM movement may hitless, the fabric changing by server power on/off are most like disruptive. In a storage world, it is a bad idea to power on / off the member frequently. if the servers are much more than 4, how to keep SAN fabric stable using today's technology?

[kerth5]

1 - Do You really know what is standby?

2 - Depends on many things, not only on vitualization. But - economic question - Have You seen, how many servers were virtualized? Imagine it on hardware servers...

3 - Vmware does not upgrade cluster on the fly, All the 4 servers are already in a cluster, You don't change config or depencities.

4 - Storage fabric will not see much. Just patches to some wwn's will be not used.

Have You tried to use VMware?

[ThePinxu]

3. Safety and management question:

Is automatically power on/off server in a datacenter a good idea? What VMware does is like an cluster upgrade on-the-fly. Even with qualified sys-admin, the best practice is those changes should be done in mid-night, not 9:00AM in the morning, and 11:00 at peak hours. VMware DRS asks us to make some dramatic changes in management polices.

[ThePinxu]

1. Hardware question:

What is the impact for hardware mortality rate because of frequent power on and power off ?

2. Economic question:

In a 4 server case, what is real OPEX when considering power cost saving and hardware maintain increase ?

[lifeintoronto]

awesome video

[VMwareFusionTV]

What a killer video. I think they're stealing from the Fusion team's playbook... ; )

[jasonboche]

Artist and song title please?

[ampads]

Franz Ferdinand - The Fallen (Justice Remix)

[jasonboche]

Thanks!  How did you know that?!

[eArtrash]

It's called

Franz Ferdinand - The Fallen (Ruined by Justice)

[bowulf]

Excellent video, now if only VMware would get this feature out of beta or experimental support. I trust my production clusters to an experimental service.