AWS will be the last big cloud to add Skylake as Azure turns 'em on

Microsoft has announced its first Azure instances running Intel's Skylake silicon, a move that means Amazon Web Services will be the last of the big four clouds to run Intel's latest silicon. Amazon announced it would use Skylake last November 2016 at its re:Invent gabfest. Google offered Skylake in June 2017, IBM switched it …

[Anonymous Coward]

>Microsoft has announced its first Azure instances running Intel's Skylake silicon

If your cloud customer has taken an interest in what silicon you're running, they are doing it wrong.

[Aitor 1]

Re: >Microsoft has announced its first Azure instances running Intel's Skylake silicon

yep, they should be using vcores etc, and tne users sbould care about service, not iron.

[TheVogon]

Re: >Microsoft has announced its first Azure instances running Intel's Skylake silicon

"If your cloud customer has taken an interest in what silicon you're running, they are doing it wrong."

Only if you have a very limited and small scale understanding of cloud. The performance of a vcore can vary significantly between CPUs and the maximum scalability of an instance varies too. And it also impacts available IOPs on clouds like Azure and AWS as max number of disks is related to #CPU cores.

Someone with an enterprise understanding of cloud will look at both the type of underlying hardware and the #vcpus per VM to optimise cost versus performance....

[Wowbagger42]

Re: >Microsoft has announced its first Azure instances running Intel's Skylake silicon

Prostetnic TheVogon is right. Whenever our on-prem customers request provisioning & sizing parameters we start by asking what type of CPU they're currently running because that will impact significantly the response time you will get (we sell calculation intensive based software) and in 75% of the cases the customers has no idea. Most never really pay attention to the base freq / turbo freq / L2 cache etc and say something like "we've got 16 core CPU's" and think that by adding lots of vCPU's to a virtual server it will perform better while in fact it's really easy to waste 30-40% of overall CPU time just by over allocating vCPU's.

So yes, it really matters what type of CPU you are using for what type of workload.

[paul2718]

AVX-512 may be useful to some.

Paul

[DontFeedTheTrolls]

Whoop-d-do!

Isn't the whole point of the cloud that you don't care what is running, its somebody else's computer and problem, you just want to know how much work you get done for a given cost.

[TheVogon]

"you just want to know how much work you get done for a given cost."

Which depends rather a lot on the underlying hardware you choose and how well it is optimised for your requirements.

[gudguy1a]

Hardware and services

Yes, good one "TheVogon". It does matter to some businesses to know what services they can run in cloud AS WELL AS the type of hardware that cloud provider is providing. Before anyone comments on this, i.e. some software is optimized to run more efficiently on certain hardware. Just the way businesses, to this point anyway, is being done. Not all software runs the same on all hardware. If only it were that easy - it would make life MUCH better for all of us.

[Anonymous Coward]

But what is the real story?

While MS moving to Skylake before AWS is interesting, I wonder why AWS has chosen to wait?

As you mention, AWS are cloud pioneers and the big cloud providers have a reputation for getting early access to new hardware to play with before it's available to the rest of the market.

So are AWS being punished by Intel preventing the use of the new chips or is AWS just unimpressed with what Intel has to offer at present? Or did AWS use up their brownie points on the last chip launch and now have to wait patiently.