What’s going on then – removing our dependency on Nathaniel

martin237 | November 2, 2011 | 0 Comments

I’ve been reading up and discussing changes in the industry recently with colleagues, this has meant a bit of a break from my beloved Bladewatch blog. So what’s going on, a lot of things really, there seems to be a state of flux of parallel running across sectors and industries with common themes in two main opposing movements.

Commoditize and reduce complexity which allows us to concentrate on really adding value where it counts

Build best in class solutions based on the business need with a focus on delivery and through that illustrating value

Both can co-exist. Both are equally valid, but the commodity discussion is compelling and equally troublesome for colleagues across the technology stack. There are three main problems, ‘the business’ want to remove delays to excellence, reduce costs and are approaching technology as a service. With the technology as a service model, I’m not interested in the detail to the nth degree, I do not wish to discuss which model of server it is and what version of Windows is going to be used I want a Risk Analytics platform and I wand it next Tuesday with these operational constraints (budget/application code compatibility for example).

This results in an interesting problem. For years we have been telling technologists to be customer facing, to identify the requirements, to offer choice and therefore empower the business, an admirable concept but less so as you scale up the enterprise, with a global enterprise with operational complexity, introducing choice even with a business aligned best intention can create opportunity for debate and discussion. It also leads to that strange situation, “I want a server”, ok what would you like, “well what’s the standard, just give me a server”, ok but what OS, how much memory, how much storage? “the standard”.

Therefore the prevailing wind for the moment seems to be moving to commodity. Commodity can be perceived as an emotive subject, ‘those architect boys’ in their design studios mandating standards that might not be appropriate but what I am talking about is a rationalisation of the standard tools covering the hardware, the os, the middleware and database through to the applications. A scenario in which packaging, and common deployment technologies can be used for automated delivery, decommission and re-deployment on demand can be defined and implemented. A situation where as a user I can say I wanted a Tier 1 web farm and like the ever so impressive BladeSystems Matrix presentation shows us, I press submit and it knows it’s going to build me a load balanced web farm comprised of web, application and database using the capacity and standards already defined in templates.

So what do I mean when I talk about commodity and rationalisation. Colleagues differ wildly when we talk of it, but let me summarise:

Server – three platforms (Unix/Linux, Virtual and Windows) / three blade and three rack server configurations (typically dependent on complexity involved)

Operating system – a common operating system offering comprised of UNIX, Linux and Windows platforms, Windows 2008 R2 Standard and Enterprise for example

Middleware – common middleware applications with specific versions per platform which change with the platform

Application – common applications for monitoring, database and solutions

Network – common defined network standards based on standard offerings, 1GB Ethernet for Virtual machines, teamed network cards for physical Windows servers, heartbeat for clusters etc

Storage – common standards combining the most effective forms of de-duplication, tiering and marginal cost – sometimes buying the cheaper storage costs the same as the expensive equivalent

Operational standards – production tiers and understood configurations for optimal resilience

Investment – this is often the emotive topic when speaking with management – this involves the concept of pre-provisioning and investing to avoid those common scenarios in which we are ‘always running out of storage’ or facing those unexpected infrastructure type issues half way through a project or as a result of a business requirement.

The fundamental principles as follows:

The infrastructure is commodity and therefore abstracted from the underlying hardware
The application never gets embedded and locked into the infrastructure because it is economically, operational and from a revenue/cost base inefficient to do so
The infrastructure is funded, cross charged and operated on the basis of lowest operational cost – it is 9 times out of 10 quicker and easier to replace a desktop than it is to have an engineer sit and discuss the issue with the vendor and order new parts to do so. The same can be said of the server. Do we want to be paying £800/£1200 per server per year for hardware maintenance or increasing our spend on warranty to 24/7 4 hour response and replacing the x86 commodity servers every three years so that we do not need a hardware support contract?
Constant analysis so that the capacity is being optimized, that I deploy a new platform on new x86 servers in six months if the utilization is remaining constant at 17%, we virtualize and re-allocate the hardware to the next project, and do the same analysis to it in six months. Two streams, one translating business needs into delivered projects and solutions, the other optimizing what we have and allocating capacity on demand in line with the business need.
Taking a high level and low level analysis. Where is it as CIO we need to be (agile, low cost/best in class), combined with where are our core spends, what percentage of the budget is on legacy platforms, what are the quick wins, what is a standard offering and why is it set that way, what operational decisions and rules of engagement have been set. Is Nathaniel walking around the data center reclaiming and re-deploying DL380 G3′s because he’s been told that there’s no budget eating my data center space/power, my hardware support contract budget and resulting in outages for the sake of £2500?

Therefore a change of focus to:

Delivery on time
Delivery to budget
Delivered once. Configured once. Templated once. and redeployable on demand.
Right sized infrastructure with scalability, reliability and best in class out of the box.
A change from component thinking to big picture thinking – standard offerings, common configurations, interoperability, reduced complexity resulting in best in class performance, operational cost and agility.

It’s not that we don’t want Nathaniel being technical. It’s that we want Nathaniel being technical with the department, the business and the enterprise in mind, not the driver, the firmware, the hypervisor version or platform.