Original URL: http://www.theregister.co.uk/2013/07/12/smes_need_enterprise_class_systems_too/

Why I'm sick of the new 'digital divide' between SMEs and the big boys

IT design by white paper is a stick to beat the little guy

By Trevor Pott

Posted in Data Centre, 12th July 2013 11:28 GMT

Sysadmin blog Recently I have been spending most of my time with enterprise CIOs and vendors' "product owners"; their dismissiveness of the needs of small and medium-sized businesses has finally got to me.

As a rule, talking about the needs of corner cases is boring to the masses, but today I just need to get up on my soapbox and talk a little about the little guy.

As a sysadmin, I live and work on the edge of the bell curve: SMEs with enterprise-class requirements but no possible means of obtaining enterprise-class budgets.

These companies seek me out to help them design, implement and manage solutions that are an order of magnitude cheaper than you'll find in any technology white paper.

As always, cost cutting comes with a price, but SMEs generally choose different compromises than mid-market organisations and enterprises.

SMEs don't exist in the same universe as hyperscale operators; it's one of the reasons why strategies such as Microsoft's Azure <--> Server "rapid release" design feedback loop do not fill us with warm fuzzies.

Microsoft's new design concept - one that many vendors are mimicking – is where Redmond dogfoods its own operating system and server apps by using it to underpin its Azure cloud.

Lessons learned there are fed back into the design for the next generation of Windows Server, SQL, Exchange and so forth. Similarly, anything the server teams come up with or neat ideas they get from partners and customers are then tried out on Azure with the next release.

That's an awesome way to build an operating system designed to run a hyperscale datacenter. Of course what the people playing with clouds get up to is an irrelevant thought to those of us way down here working on our local ant colony.

Disconnected in approach

The options facing my clients are often:

When I talk to my peers who work in "proper" IT they tell me that these options are the only options that my clients should consider. I tend to disagree.

We live in a world where fickle consumers spoiled for choice can and will pack up and go elsewhere if you are down for a few minutes, let alone hours at a time.

Larger companies and governments sit on payments to suppliers far longer than they should, draining capital from smaller organisations. Meanwhile, the commoditisation of everything under the sun drives margins into the ground.

Buying a single "proper" enterprise server, filer, switch or whatever can be gambling your entire company on not becoming an insolvency statistic. That "four-hour enterprise service window" can – if the outage hits at the wrong time of year – be enough to sink a company.

No matter how unlikely the risk that the device in question will break, SMEs simply cannot afford to be a statistic.

This is what I bear in mind when I do my job. Every single time I design a system I think about a friend of mine, a single mother with two deaf children who works herself to the bone to give these kids a future.

If I design the wrong system or make the wrong choices then the company she works for will take a massive blow. Maybe her job will be cut – maybe the company will fold – and where will she (and her kids) be then? Where will all those people I've worked with for all those years be?

Vendors say buy

For many in our industry, "competence" is synonymous with "design by white paper". This word is wielded by CCNAs (Cisco Certified Network Associates) foisting $3,000 Cisco routers on SMBs with $60,000 yearly revenue and armchair systems architects who cannot disconnect from their lifetime's immersion in Tier 1 services and solutions.

"Competence" in our industry has become synonymous with regurgitating corporate messaging and deploying the newest solutions instead of those most fit for purpose.

It is both word and weapon: a sword oft drawn to mock and ridicule those who cannot afford to spend three years' gross income on the white paper solution to a problem.

I have had technology journalists at other publications belittle me for talking about such things. They believe that edge cases get too much press as it is.

We should focus on the middle of the curve, I am told; it is our ethical duty to use our privileged position of textual influence to spread "proper" administration and design principles.

This cry has been echoed by many a comment-poster as well: push the "right" way to do things, talk about the "right" vendors and only discuss the "best" technologies, means of application and most common use cases.

First, we assume a spherical cow

I acknowledge that this is probably a logical fallacy of some variety, but I can't get past the concept that more people would focus on "design by white paper" if it were such a great thing that worked for everyone.

The fact that discussions about all these edge cases, startup vendors, open source, cheaper alternatives and novel design approaches keep getting page views would seem to indicate to me that there is a demand for that information.

That this demand exists is anathema to many. To not champion the new simply because it is new is to hold back progress. We are to eat what's put on our plate, pay what we're told to and accept what works in overly simplified simulations of what our use cases "should be".

I posit that a new digital divide is growing. This one is a gap of abstraction. Those with large enough budgets can assume a spherical cow exists and proceed from there to design their entire farm.

Where the world doesn't meet their models they can just toss some more money at it and the problem goes away; the entire nature of their design process allows it.

Virtual networks and virtual storage running virtual desktops and automatically deployed applications in virtual data centres running on clouds of hypervisors managed by statistical modelling and predictive analytics tools. Everything can be solved with a few more nodes or a few more instances.

None of that works in a world where the cost of a single Enterprise CAL (client access licence) is a major expense, to say nothing of the server licence or the cost of the actual hardware.

So what is "proper" IT design? Is it an addiction to off-the-shelf technologies and white papers, compiling everything from source and knowing how to write machine code or is it far more complicated than either?

I believe that "proper" IT is about matching the solution to the problem and that it rarely ends up being the same exact solution twice. Your mileage may vary. Debate, as always, in the comments. ®