Power laws & the enterprise application development interdependency network

November 10, 2008 – 8:50 pm

Shared Context recently completed an analysis of a Fortune 500 software development interdependency network.  The network was composed of approximately 30 strategic applications.  Dependencies between the versions of each of these applications in various environments (Development, Integration, QA and Production) were captured over a three month period.  The dependency distribution followed a power law.  

Future posts will describe what this means and why it is important.  Other relevant properties such as attack tolerance, SLA adherence, project planning, team dynamics and design churn will also be discussed.

Sphere: Related Content

Release Management and the Giant Component

June 19, 2008 – 10:02 pm

Imagine an organization with many independent distributed applications, developed by bounded teams that have little or no interaction with each other. Each team moves its application into the production environment according to its own timeline. Over time new applications are built that need to access existing applications or other new applications. As more and more applications are developed and linked something suddenly happens; they are all connected and a giant component is formed. No longer can each application move to production on its own. No longer can each team make decisions that apply only to it or its nearest neighbors. Suddenly everything has to move at once. If you know your application and project interdependencies, you should be able to see this coming. If you don’t, many individual timelines will be missed and you may find your information systems department has no team in place to manage this emergent system. Here are some suggestions:

  • Map project interdependencies at a high level.
  • Map application interdependencies at a high level (which app calls which app).
  • Develop a means to traverse and analyze the interdependencies.
  • Identify the formation of smaller clusters. These clusters will often have to move to production together.
  • Check your strategic plans against existing interdependencies. This will allow you to predict the formation of new clusters and potentially the formation of a giant component.
    Sphere: Related Content

Electoral dynamics in the 2008 presidential primaries

February 13, 2008 – 10:59 am

One of the UW-Madison’s best kept secrets is a weekly seminar on chaos and complex systems held in the bowels of Chamberlin Hall. Yesterday, Charles Franklin from the UW Department of Political Science gave a timely presentation on the electoral dynamics in the 2008 presidential primaries. The data he presented show just how well candidate strategy played / plays in the face of real world electoral dynamics.

Sphere: Related Content

A Multiscale Snow Day

February 6, 2008 – 5:46 pm

Thanks to a shield of intense snowfall bringing 8-12 inches of snow, schools are closed and I have the unique opportunity to stay home. Between episodes of playing with my children, I am able to pick up development of a program that runs a simple multiscale analysis (see this, this and this) on project interdependency data.

The goal of the first iteration is to capture and analyze data for large scale development efforts that are composed of multiple teams with independent timelines, clients and mixed levels of expertise. Much of the work that keeps Shared Context in the black comes from planning software release efforts (Release Management). Success is often judged by meeting timelines and protecting production environments. Organizations, especially large organizations, typically have mixed success in this area. This is due in part to the issue of applying linear project management “best practices” to projects that display characteristic non-linear complex behavior. I in no way believe that an application can manage this complex behavior but the simple act of showing a system’s dynamics can really help an organization to see that their approach may need reconsideration. If nothing else, it uncovers the most significant areas of misalignment.

Sphere: Related Content

HBR meets Complexity

November 20, 2007 – 11:09 am

November’s Harvard Business Review visits CognitiveEdge’s Cynefin decision making framework. Feedback from clients who read the article is positive leading me to believe that this is the most digestible article related to complexity in quite some time.

Sphere: Related Content

Complexity Related Domains

November 1, 2007 – 5:30 pm

Work in complex systems science/thinking/engineering carries with it an understanding that our attempts to control are of limited efficacy. Nonetheless, we continue to try and that act of trying in itself is important, however fruitful the final outcome is when compared to the initial intent. In an admitted effort to control the inevitable sequestration of domain names by resellers, the following domains now point to sharedcontext.com:

  • complexsystemsengineering.com
  • complexsystemsengineering.com
  • engineeredemergence.com
  • multiscaleanalysis.com
  • multiscaleengineering.com
  • complexengineeredsystems.com

Anyone working on a project (preferably of an academic nature) actually related to these concepts is welcome to contact us to discuss their transfer either as a donation or at the cost we paid to register them.

Sphere: Related Content

ICCS Day One

October 28, 2007 – 11:27 pm

Today was the first day of the ICCS conference. The pedagogical sessions started the conference on a positive note, literally, as Dianna Dabby played generated variations of Bach’s Prelude in C. Utilizing Lorenz Equations and a fairly elaborate set of rules, the generated variations were actually pleasant. Hats off to musical complexity. Ed Fredkin followed with a very interesting presentation of his work on Salt, a computer that actually enforces conservation laws (think Computational Physics). After Ed, Gyan Bhanot gave an amazing presentation on the history of human migration based on Evolutionary Genetics. Did you know we have a common mother and she lived in Africa approximately 206,000 years ago? Long live mitochondria, at least in females. And long live Numerical Methods, particle based approximations to solids, the Material Point Method, Navier Stokes Equations, Adaptive Mesh Refinement, and a whole slew of etc., if you want to know how to simulate fires and explosions. That is exactly what Martin Berzins wants to know. Given the fact that we have high explosives traveling, and sometimes exploding, on our roads, we can all rest easier knowing he is on the job. And thanks to Liz Bradley’s work on non-linear computers, the engines in the trucks pulling those high explosives may soon benefit from her ability to utilize initial conditions to control longer term behavior. Evelyn Fox Keller read her work on function, purpose and agency. Frannie Leautier picked things up with a very elegant discussion on how complex systems reasoning can help the challenges of sustainable development. Many, many challenges remain, and much work awaits any eager researcher in the area of sustainable development. Photosynth’s multiscale visualization, created in part by Blaise Aguera Y Arcas, was phenominal. The implications of the tool’s potential to model large parts of the world from unstructured data, when combined with temporal dimensions, are immense. The day was closed by Barbara Jasny, the Supervisory Senior Editor of Science Magazine who was able to give convincing answers to many skeptical questions about how articles are chosen. The day of the print journal is not over, nor is the frustration that many feel when their work is excluded.

Sphere: Related Content

Complex-System Engineering Recommended as System Engineering’s Second Branch

October 12, 2007 – 2:05 pm

The abstract for Michael Kuras’ upcoming presentation at the ICCS conference represents a well needed recommendation for the Systems Engineering discipline. That being said, the concept should reach beyond Systems Engineering and into what could be called everyday engineering. The Global 2000 should take note.

Related: A Multi Scale Definition of A System - M. L. Kuras

Sphere: Related Content

7th International Conference on Complex Systems

October 8, 2007 – 9:08 pm

The New England Complex Systems Institute will host the 7th International Conference on Complex Systems October 28th through November 2nd. The Systems Engineering of Complex Systems Conference, hosted jointly by NECSI and the International Council on Systems Engineering, will be held in association with the ICCS. The fact that these two conferences will be held together is significant in that it demonstrates a movement to integrate the more traditional Systems Engineering with the more forward looking Complex Systems Engineering.

Sphere: Related Content

An Implication of Web 2.0 Interdependence on Business Activities

July 23, 2007 – 2:23 pm

The excitement surrounding the integration of the enterprise with Web 2.0 is hype-rich and is the corporate world’s long-delayed introduction to systems thinking. While the possibilities for corporate growth are significant, the implications of interconnected and integrated business processes may soon cause businesses to have to rethink their current concept of a business activity.

Understanding Implications

We often present the concepts of initial conditions and conditions at present (CAP) as a means of understanding and managing systems. During one session with technical leads and project managers from various internal development teams we presented the question; “Would development grind to a halt if everyone knew the implications of their changes?”. The question was followed by a nervous laugh. No one could definitively answer the question. They did know that given their experience and the significance of their development efforts, the implications of their changes were well beyond their ability to predict. The nervous laugh really came from the understanding that accurately making those predictions was part of their charge.
For some types of systems, one could potentially know the state of a system in the future if all of the initial conditions are known and the system is free from external influence or random noise. The problem of course is that we never really know all of the initial conditions or escape the reality of external influence and random noise.

An Implication of Integration

Businesses often have the luxury of processing their activities within the confines of their organizational boundaries. There are many external influences but such influences can often be avoided or delayed because the strength of the connection between the business process and the external influence is typically not immediately evident or enforced.

Now lets think about a business that has its processes connected to external processes maintained by service vendors, governments and other applicable entities. These relationships (connections) are now stronger. A change to one can have an immediate impact on another which can thus provide immediate feedback as to the efficacy, legality or feasibility of that change. While this is often considered a net positive it has many implications. One of these implications is greater insight into CAP conditions which in turn allows for better prediction. We all applaud this because of our desire (and charge) to control. But therein lies a problem. Knowing more about the actual (or perceived) implications of a change may mean that many current and planned business processes are found to be unfeasible or have undesirable results. Many businesses succeed because their processes are feasible within certain boundaries. External implications are often ignored or unknown, allowing a business to operate for a period of time without consideration for the impact of its processes. Interconnectedness may cause many businesses to have to face more of the implications of their activities and thus potentially significantly modify their behavior. This will result in costly changes for business, regulatory ambiguity and many new opportunities.

Sphere: Related Content