ESX 3.5 vs 3.02

[RAlltuck]

What will be some of the major advantages of the upcoming ESX 3.5 over the existing 3.02 version?

[Michelle_Laveri]

You might want to look at the very brief summary I wrote on my blog

http://www.rtfm-ed.co.uk/?p=452

Which I more or less pulled from the announcement of the new features last week:

http://www.vmware.com/company/news/releases/esx_35.html

http://www.vmware.com/products/vi/features.html

http://www.vmware.com/products/vi/whatsnew.html

Regards

Mike

[RoblLaw80]

So if I'm reading this correctly, with VMware Infrastructure 3 Standard High Availability Acceleration Kit I'd be able to set up two 2-processor ESX servers, the Virtual Center server, then additionally buy VMotion, and I'd be good to go with a fail-over redundancy configuration?

[Dave_Mishchenko]

Vmotion will allow you to move a VM from host to host while it is running and isn't required for the HA option. The HA option will just restart the VM after the host fails so the VM boots from scratch as if you had pulled the power plug. You don't require vmotion licenses for HA to work.

[RoblLaw80]

Ahh, for some reason I had the impression that VMotion had something to do with if a server dies, it automatically moves it to the other machine.

[Bwhite]

New capabilities:

- VMware Storage VMotion - live migration of virtual machine disks from one storage system to another with no disruption or downtime

- VMware Update Manager - automates patch and update management for VMware ESX Server hosts and virtual machines

-- Integration with VMware Distributed Resource Scheduler (DRS) enables zero-downtime VMware ESX Server host patching capabilities

New management solutions:

- VMware Distributed Power Management (DPM) - automatically powers off and on servers as demand for compute resources changes

- VMware Guided Consolidation - discovers physical servers, identifies consolidation candidates, converts them to virtual machines

All very cools stuff

[RAlltuck]

All the new features look/sound great (not sure I would use some of them).

Is ESX 3.5 version suppose to be any faster than 3.0.2? I thought I heard a sales rep say it was 40% faster.

[Dave_Mishchenko · Accepted solution]

It may be this reference in which case the 40% is due to a number of other things as well.

Dell said that with VMware's ESX Server 3i hypervisor embedded in flash rather than residing on a hard disk, and with the doubled memory capabilities and I/O upgrades, the Veso server will have about 40 percent better performance than similar virtualization servers and will consume about 25 percent less power.

http://www.data-storage-today.com/story.xhtml?story_id=011000VUGGLO

[larstr]

I'll believe those numbers when I see them for real. The virtualization overhead isn't normally as large as 40% unless you're doing something very unusual (like iometer max throughput tests).

Lars

[RParker]

Sales guys have a tendency to over promise and under deliver, and they never talk to the REAL techs behind the scenes, they just care about $$$$.

If you asked them what the differences were, or even how ESX works I doubt they could tell you, they just know to basically animate a brochure. Their job is to get you interested and get the word out.

It's your job to research and determine if ESX will meet your needs, by reading and getting all the information.

[RParker]

yeah, I am with you. That sounds very suspicious, I would expect a slight increase, maybe even very nominal, but 40%! I doubt it.. and I would LOVE to believe that.

[wila]

How about the logic, remove a disk and add flash, save 25% energy.. ? Sure you're using less memory and your disk might use a couple of watts as well, but it's not 25% of a server. These figures sounds severely marketing rigged.

I guess they are saying that because you can now address more memory you can virtualize more VMs on one server before running out of memory, hereby being able to save on energy.

The harder a marketing person lies, the less likely us techies will believe them next time.

[ChrisYearsley]

Yes, my understanding is that Dell are in this case saying "performance = max hosting capacity of server". Therefore just by adding extra memory slots they've increased "performance". The Dell VESO server is now not due until next year, so it will be a while before we find out for sure!

--

Chris

[Anders]

There 40% reference was made on stage at VMworld and I believe it was regarding 10g servers vs 9th generation Dell server.

It was not 3.5 vs 3.0.

That said, there are some great performance improvements in 3.5 in areas where we are struggeling today.

- Anders

[williamarrata]

Hi Mike,

Would you please look at "Creation Dates of Snapshots" If there a person that could give me good information, that would be you.

Thanks

[larstr]

That might be the case for the final version. Those improvements seems

not to be present in the current beta. But I guess that's as expected.

Lars

[Anders]

In regards to 3.5 performance you mean?

Do you have access to the non-debug RC build yet?

- Anders

[larstr]

The version I tested was the one on the usb key from vmworld+updates. Storage performance was ~30% of 3.0.2. I guess that means it has debug symbols included.

Lars

[Michelle_Laveri]

Ahh, for some reason I had the impression that VMotion had something to do with if a server dies, it automatically moves it to the other machine.

Well, at VMworld the founder of VMware did show "Continious HA"... a "primary" VM ran on one ESX host, and a "Shadow" VM ran on a second ESX host. They were kept in synch by a process similiar to the expertimental "record" feature in Workstation.

Now the announcement doesn't appear to use the phrase "Continous HA"... or indicate when this feature will available...

Regards

Mike

[larstr]

Watching the latest webex[/url] confirms that Continuous HA wont be available for 3.5. The HA in 3.5 will however be somewhat improved as it now also will be able to detect OS failures.

Lars