Private clouds kinda suck, you know?

Enterprises...

They want to buy a cloud solution, and indeed sometimes do, but the organisational changes needed to actually deliver it as a service over time are often far more complex than expected (its clearly not in the sales pitch)

I disagree to an extend with the horribleness of the various solutuions described in the article as the products all have their strengths as well as weaknesses.

But, if you have a network team, a storage team, a database team, several disparate development teams and suddenly expect them to merge and cooperate it does not really matter how the infrastructure is provided, private, public or managed. they all still need to agree that the cloud will provision storage, IP networking, elastic nodes requiring applications to demobilise nodes, etc. This is the hard bit - not the management solution itself - its an adjustment of approach entirely.

OwnCloud. Open Source, runs on Linux and gives us global access to our files anywhere where there is Internet.

It isn't as 100% functional as DropBox, but fills our needs.

Shadow IT

I was doing that at the desktop many years ago with Excel/Access, when the users wanted to get something done without it being costed as a multi-million dollar project because their IT departments couldn't think in terms of smaller scale solutions. Shadow IT in the form of cloud doesn't therefore surprise me at all. What does surprise me is that any IT system allows users to get through to such public cloud environments.

Also Confused

Perhaps a definition of "Cloud" might be useful? What is it? Is it just compute power hosted someplace else? Is it software provided by someone not on premises? If it is either of these things, then businesses have been providing "private cloud" services to companies for decades. If it is something else, then I'd like to know.

When I started in IT we used a drawing of a cloud to represent networking outside of our boundary firewall. Internet, direct connect to service provider, etc. Somehow this evolved into marketing bull shit terminology to represent services.

In the end, it is robbing Peter to pay Paul. Cloud infra is no less of an expense than on prem. From an administrative standpoint all that is eliminated is patch management. From a cost standpoint, if you are hosting large amounts of servers, you can save on rack rent or power costs. If you are a smaller shop, the savings won't be there. The rest still requires hands on or scripted configuration. Either of which cost money in engineers to develop or maintain.

We recently went to O365 to "save on costs of infra and allow our team to work on other projects". This was in October. We are only just now freeing up time to work on our normal work load due to the cock up that is O365 and all the crap that doesn't work as expected. New projects? Not likely to get more room for these due to the "unexpected" overhead of managing Exchange. Still. We shifted the location, not the work.

While there is value in off prem hosting and services, I expect that the pendulum will swing back the other way in the next 5-10 years as companies realize that they still need or require on premises services and expertise (have you TRIED working with MS on service issues?). That should take me close to retirement. Then I can get a fun job like cabinet making or some other work that doesn't require me to work with idiot executives who fall for each others BS over an over again.

It's relabelling but with a twist

Think of all the work required to get an older enterprisey three-tier application up and running in a big company on physical infrastructure:

- Architecture and design phase, all following billions of pages of "best practice" bought from some consulting firm (weeks)

- Hardware specification and procurement (often takes weeks/months)

- Data center planning (usually a couple weeks to secure a few U of rack space)

- Network planning (switches, switchports, VLAN assignments, IP address assignments, firewalls)

- Storage planning (LUNs required, backup strategy, fibre channel cards and zoning if you're real oldschool)

- DBA planning (sizing, backups, selection of software)

- OS and software licensing and procurement

- Hardware reception, rack and stack

- Network team configurations

- Storage configurations

- Bickering back and forth between network, storage, DBA and infrastructure teams during integration

- OS install

- DBA wizardry

- And finally, the enterprisey app install

The difference between this and virtualization is the ability to cut out the procurement phase assuming capacity exists. This is the VMWare virtual infrastructure phase, making server deployments less painful.

The next leap for older-school enterprises is rebuilding their apps to be more cloudy, using web services and the like. This is going to be the bigger leap that makes it easier to decouple from some of the siloed IT departments. Of course, the complexity doesn't go away, but it does get abstracted.

Nothing Ventured, Nothing Gained :-)

How can all of this be if enterprises are also rife with "shadow IT"; the adoption of public cloud services undertaken by various departments specifically to avoid dealing with internal corporate IT? If both are occurring at the same time, why? ..... Trevor Potts

The one is BetaTesting New Pioneering Privateer Programming for the other, methinks, TP.

And more than a little something different for the Register to Register with the Federal Register, Trevor?

Who Dares Win Wins.

Really good article

Nice to see something that isnt just kissing upto vendors.

I agree that most of the products on the market that promise private cloud are pretty awful and Azure Stack seems to have lots of potential, well anything that could kill off scvmm in the long run is good!

"Are internal IT departments simply incompetent? Are they not delivering the workloads corporate customers are demanding, or not delivering it quickly enough?"

I think most corporate IT systems, policies, procedures, etc. are set up in such a way as to minimize the pain of the people in the IT department. Can't say that I blame them given some of the stupid users they have to put up with, but it does get in the way of getting work done once in a while.

Shadow IT and Private cloud in parallel

The article questions how it can be possible that the enterprise is focussed on building a private cloud, while the staff are also known to be performing shadow IT and procuring their own public services.

I'd say the focus on private 'cloud' is a result of the users demands... and it's a catchup.... the organisations are probably using different methods to block information leakage and use of public services through their network and internet links.

So yes.... both are absolutely happening in parallel

Enough double-speak!

"thought influencers"

Why is that word a thing? There's so many, much more appropriate ones:

Huckster, snake oil salesman, side show barker...

if you're being polite, just call them salesmen, but ask yourselves, "Why the fuck am I paying this guy to sell me shit?"

some-when???

I guess since you don't speak English and you are writing for a living, you should soon be broke. I can't take this article seriously when I read that. Isn't "sometime" a real word? I am not a native English speaker so I could be wrong.

It's very easy to understand what is going on. However, you have to understand a few things:

1. Companies (should) make money.

2. Specialization.

Because specialized software will have more butt wipes than generic software and will also have more sales wolves, sheeple are bound to follow the sales wolf.

Anyway, it doesn't take a large amount of analysis to understand that the portable music player will sell better than the PDA that can also play music because most people don't enjoy thinking. Most people enjoy feeling.

Computing devices have become mainstream, and while engineers and hackers might be annoyed by that, it does us no good to prolong the inevitable. Idiots will take over because they are good at leading other idiots. So the quicker the techopolypse ensues, the better off humans will be because the idiots will generally be eliminated.

Oh, wait a moment, I sound a bit cynical. Were we talking about rain?

It's all about API's

I think the author of this post is "bang-on", including his comments regarding OpenStack, and I run a company that monetizes OpenStack. There seem's to be some discussion with respect to cloud and what it is, why we need it, and is it really worth the effort ? I believe there is a need for an on-premise "AWS" like cloud. Yes, self-service is important, likewise with elasticity and many of the other features referred to in the comments but what makes AWS so appealing is their API's - while Microsoft and VMware have fallen very short here. OpenStack has created an abstraction layer that addresses the entire datacenter via API's. Yes, some of you will argue that VMware has API's and they do, but those API's are for front-end access for the purposes of automation, aka DevOps. OpenStack's API's are front-end and Back-end. With OpenStack, an enterprise storage array can speak to the Block API (Cinder) and front-end DevOps can automate the provisioning of storage, snapshots even replication. If you replace that storage array with another Cinder supported array, none of the DevOps front end scripts need change, because we are issuing Cinder commands not array specific commands. With Microsoft & VMware those scripts would require modification, in some cases very little, but still modifications. This same abstraction applies to Network and Compute. It's all about the API's. Assholes & Opinions, everyone has one, this is mine.

incompetence

Are internal IT departments simply incompetent?

I think you often find the issue is bit higher up.

private clouds better

Your PHB has read somewhere that clouds are the thing so if you don't call them clouds he probably won't know what you're talking about.

Are private clouds good? You bet ... unless of course you prefer you prfer to put your major data assets in someone else's hands and trust that (a) your data is safe from disaster and "leakage" and (b) when the excreta hits the ventilation equipment that the service provider agrees that your high priority is also his high priority. For most companies IT is their lifeblood; it involves every aspect of the organization and therefore how the company makes its income. Willing to cede control to someone else somewhere else? You'd have to be crazy.

No mention of Joyent?

No mention of Joyent and SmartDataCenter?

Shame, one of the easiest to manage and cost effective private/public cloud offerings.

Great Post.

HP has canned Helion which is sort of ironic as you almost gave it a thumbs up !

http://www.theregister.co.uk/2015/10/21/hps_public_cloud_gets_eol_date/

Good work, though I'm curious

Decent article there, Trev. My curiosity was piqued however by the following comment:-

"If you ask about things like self-service or API-driven workload management they'll lose their minds all over social media"

If I didn't know better, I'd say it sounds like a company that may or may not rhyme with Buttflanix.

Great article, Trevor.

here is another problem:

Networking abstraction and service insertion is an AFTERTHOUGHT. Hairpins everywhere on networking, true networking abstraction not possible and even with an SDN helping its plugins for kubernetes, openshift, docker, CNI/CNM, vpshere<>esx<>kvm is complex, openstack and SDN integration is tough- most SDNs are not in-box or batteries included with openstack .

So we have bad SDN integration. Bad integration with cloud native/containers. We have issues with having containers and VMs and physical machines talking to each other without twisted issues.

distributed firewalls, POD level network isolation, LBAAS/loadbalancing and DF/firewall , etc, simply integrating checkpoint/PA/NSX-DFW/fortinet is a HUGE pain, and then A10/F5 on that, pain.

Private cloud is so bad and now the word has started to move onto cloud native, managed services, SAAS, etc. Nobody wants to pay attention to all this trash wiring anymore.

REWRITE YOUR APPLICATIONS because trying to do things the old way on private cloud __STINKS__. Until then avoid any complexity and changes - go all in - containers and lambdas and the like. Avoid marrying private cloud anything and make sure your public cloud strategy works across clouds when possible. (Avoid too much managed service abuse at Amazon which causes lock-in ).

Its a brave new world and private cloud is lost. Its horrible.