The NoCloud Organization Part 1: Would You Like Fries With Your Cloud?
When Ray Kroc looked to expand his restaurant business beyond just a few stores and into a larger franchise he would face several challenges in delivering a consistent standardized product on a larger scale. One challenge was labor — if the restaurant was to expand under the franchise model, it would be necessary to rapidly find labor replacements at low cost and minimal training. How would his restaurant be able to acquire new labor at low cost and yet still be able to train them to provide a highly consistent and standardized product?
The solution was to turn the kitchen into something that mimicked a Model-T assembly line. New tools, such as larger grills and condiment dispensers would be to create a consistent and easily repeatable process. In addition the division of labor would be split across key inputs such as the hamburger patties, condiments and toasting buns such that each employee had only one simple repetitive task to do. Not only did this provide consistency, but it reduced the level of training required such that new employees could quickly be plugged into the hamburger assembly line in a business where high turnover was common.
The short 30 second video below highlights some of this:
In taking this approach Ray Kroc was able to reduce the time it took to provision a hamburger from 20 minutes to 30 seconds. New labor could be quickly plugged in with little training and at low cost– one would focus exclusively on condiments, and another buns — while the product became consistent and standardized.
CAN THIS APPROACH WORK IN IT?
I was working at a large global company which had chosen to standardize on IT products from a single vendor wherever possible. Part of the reasoning behind this was a belief that the standardization of contracts and support would offer benefits, as well as a belief in greater interoperability coming from a homogenous environment.
For one of my projects I was looking into a solution that I believed had the potential to offer more value than what was available from this single vendor, and eventually they were able to get an audience with our Vice President (who was based in another region). The vendor would call me back the same day to tell me that during the meeting the VP had indicated that he wanted to make IT tasks “so simple that a monkey could do it” to explain why he wanted to stick with a single vendor.
Indeed at this organization, people were corralled into a strict division of labor wherever possible, and tasks were made to be as simple as possible (or as simple as management believed they were). This is a significant contrast to picking best of breed solutions and demonstrating faith in your engineers to design and implement the best solutions. Can such an approach be effective knowing just how complex cloud computing can be? How do you think this approach worked out for this organization? (Check back later for Part 2 of this article for the answer).
ARE WE BUILDING A HAMBURGER?
Perhaps in the 1990’s it was possible (to some extent) to silo your infrastructure into repeatable and standardized tasks, but as we move towards IaaS and PaaS models things are a bit more complex. There’s only one hamburger on the menu, but your organization’s provisioning menu could have many choices, each with different operating systems, hardware allocations, applications (including n-tier), and even performance and security profiles. And with new virtualization and automation layers, the core elements are far more inter-dependent and intertwined than they used to be, making a strict division of labor a challenge in many areas.
Now think about all the IT acquisitions, mergers, alliances, startups, and new products revolving around the cloud. Can you really buy from just one company? Are you really making things easier for your “monkeys”, or are the “monkeys” now pulling on levers that are just grinding the gears of your IT machinery?
Having said that, some are trying to bring best of breed solutions to market under a single support umbrella such as VCE’s Vblock. The Vblock (discussed here) is VMware software, Cisco hardware, EMC storage plus orchestration software all designed to work together to provide an efficient infrastructure. While there are significant advantages to converged infrastructure (includinging consolidated purchasing, contract and support), you can no longer continue to effectively manage IT as separate silos. There will still be deep SMEs (subject matter experts) in many core areas, but you will need skilled generalists (as Nick Weaver explains here) to be able to transcend across these areas (and teams) to support, develop and automate the solution. If your network people aren’t talking to your virtualization people, who aren’t talking to your storage people, your operations probably aren’t going as smoothly as they could be.
A DIFFERENT KIND OF COMPLEX SYSTEM
Recently I had the opportunity to observe medical teams in close detail. I would participate in rounds and had numerous impromptu discussions with medical professionals on everything from medical technology to specific patient cases. Each medical professional would have a specialty or area of expertise including pulmonary, cardiology, general surgery, plastic surgery, pain management, physical therapy, nutrition, infectious control and more. While each had a formal role in the organization, they all needed to have enough knowledge of medicine to see “the big picture” and to discuss with each other about how to best approach a given case. There would be countless trade-offs between the different medical “silos” that would need to be considered and understood by all. Some of the goals of pulmonary might conflict with goals from the pain management and surgical teams and there would be numerous combinations of such trade-offs and competing metrics and goals. There was no room for egos — no need to prove anything and no agendas to be driven home. Just professionals with differing areas of expertise transcending their own areas and comfort zones in order to discuss a case and seek the best approach for a patient. If any organizational and/or cultural walls would interfere with this ability to collaborate across different medical areas, the patient might be adversely affected.
Cloud computing — which is really what today’s “computing” wants to be when it grows up — is far more complex than 1990’s style IT management. There were always the silos of compute, network, storage and applications, but the interactions are far more complex today, not unlike the systems within the human body. To give one example of just how complex cloud computing is, Christian Reilly writes “today, there isn’t a CMBD tool on earth (yet) that can realistically and efficiently keep pace with the inherent fluidity, agility and flexibility of even the most well intended cloud deployments.” Or in other words, we’re not provisioning hamburgers here.
It sounds a bit cliche — perhaps because it is said so much — but the biggest obstacles today to the benefits of cloud computing are not technology, but people. Mind sets, cultures, organizational models and operational processes can each be major barriers to cloud computing. While the goal of cloud computing may be to operate with increased automation and efficiency, a model based on single vendors, monkeys pulling levers, and assembly-line style silos may not be the best approach. In Part 2 of this series, we will take a look at an organization which might just have done all the wrong things when it comes to effective IT in the era of cloud computing.