What does an enterprise cloud look like?

In late 2014 I wrote about Software Defined Infrastructure (SDI). I revisited this early last year. This year I expect the first mainstream SDI blocks to emerge, likely under the moniker "Enterprise Cloud". So what does the enterprise cloud of 2017 look like? A number of players are entering, or have already entered the …

[TRT]

I'd say...

a bit like this.

[returnofthemus]

Re: I'd say...

Thanks, but we already have an Enterprise Cloud

[Anonymous Coward]

"The public cloud has also helped drive a stake through the heart of the antiquated notion of dedicated nodes or clusters for single workload types."

This isn't strictly true. If anything investment in single-purpose platforms is well up in big organisations, as data-intensive workloads on virtualised kit run like a fat, two-legged dog. Even moderately sized parallel/distributed workloads can bring a top of the line SAN to its knees, resulting in disk wait times on the order of seconds. Virtualised network storage is just not optimised for the bulk-parallel sequential IO workload.

You end up needing dedicated DAS, which implies you're exposing the internals of the machine to the consumer, breaking the SDI/SDN model entirely. Once you're at the point of having specific machine types for specific workloads you're effectively back at dedicated infrastructure, so if you're at the right scale you may as well cut out the 20-30% performance overhead of OpenStack and just buy real kit.

A similar constraint applies to GPU-accelerated computations, which are increasingly common.

The only reason either of these workloads work well on the public cloud is because you can provision many, many much smaller machines than you otherwise would to leverage the near-unlimited bandwidth promised by the public object stores. Show me an on-prem storage system providing the same kind of global access/global bandwidth as S3/ADLS etc. and then we can start talking about the end of dedicated kit for dedicated workloads.

[Anonymous Coward]

You're completely wrong. And the assumption that every single workload fits cloud use case is wrong. Typically most demanding apps run on bare metal, not even on virtualisation platform. Speaking about SAN and storage performance since very long time OpenStack supports multiple storage backends. For the most demanding workloads typically one use dedicated all flash Ceph cluster or alternatively all flash SAN. You may also use ephemeral SSD disks on compute nodes. So there are plenty of options available for variety of workloads and OpenStack provides lot of flexibility in this area.

[Anonymous Coward]

"For the most demanding workloads typically one use dedicated all flash Ceph cluster or alternatively all flash SAN. You may also use ephemeral SSD disks on compute nodes"

If you read very, very carefully that's *exactly* what I said.

[John Smith 19]

Good to see a report that says what's not there yet, rather than just what is.

I'm always suspicious of any presentation that denies they have any loopholes in their product.

I guess the message is "A lot better than it was and mostly there, but still some parts missing."

[J. Cook]

+ MANY on the 'automated backup' solution.

*goes back to the rock face to continue fighting with current backup 'solution' which is stupidly complicated for the simple act of getting backups of a friggen file server. *

[mr_souter_Working]

yeah - not so much

As someone who is working for one of those global organisations that you would expect (or at least hope) to be leading the field in this sort of stuff (our advertising crap says we are anyway) - I can tell you that we can't even get 5 year old tech working properly for our internal systems. The last excuse I was given was - "not enough storage allocated to enable document versioning - and not enough storage available to assign more".

All in all, I will believe it when I see it.

[Alan Sharkey]

It's part of the solution

I think this article (and all of Trevors) are very interesting, but they all miss the plot a bit.

From a BUSINESS perspective, what they care about is IT as a Service. They are not interested in how it happens - you could have 41 monkeys working in a coalmine for all they care - as long as the result enables a more productive business.

This new paradym we are entering (SDI etc.) is just another in a long chain of "enhancements" that may or may not meet a business need. It's a techie solution, not a business one. When you add in all the other BUSINESS contraints, this a s small part of th eoverall solution.

I like to think I am an IT "architect" who understands not only the the technology which is ongoing upcomng and relevant (or not) but also what the business needs to survive and grow. I look at all this new fangled "stuff", and then try and see it from an overall perspective - how will this make my customers' (yes, I am a consultant) business grow? It may reduce costs in the long term, but what we see is that money is spent on the next "innovation", it lasts just about lomg enough to recoup the spend and then the next innovation is upon us. Has it helped the customer grow their business or do they just see IT spend growing and creaming off the top 1-5% of profits?

On the other hand - it's great fun playing around with it all.

ALan

[one crazy media]

Bunch corporate honcho's running around clueless and forcing the word CLOUD in every sentence