Millions of people experienced the frustration and failures of the Obamacare website when it first launched. Because the code for the back end is not open source, the exact technicalities of the initial failings are tricky to determine. Many curious programmers and web designers have had time to examine the open source coding on the front end, however, leading to reasonable conclusions about the nature of the overall difficulties.

Lack of End to End Collaboration
The website was developed with multiple contractors for the front-end and back-end functions. The site also needed to be integrated with insurance companies, IRS servers, Homeland Security servers, and the Department of Veterans Affairs, all of whom had their own legacy systems. The large number of parties involved and the complex nature of the various components naturally complicated the testing and integration of each portion of the project.

The errors displayed, and occasionally the lack thereof, indicated an absence of coordination between the parties developing the separate components. A failed sign up attempt, for instance, often resulted in a page that displayed the header but had no content or failure message. A look at end user requests revealed that the database was unavailable. Clearly, the coding for the front end did not include errors for failures on the back end.

Bloat and the Abundance of Minor Issues
Obviously, numerous bugs were also an issue. The system required users to create passwords that included numbers, for example, but failed to disclose that on the form and in subsequent failure messages, leaving users baffled. In another issue, one of the pages intended to ask users to please wait or call instead, but the message and the phone information were accidentally commented out in the code.

While the front-end design has been cleared of blame for the most serious failures, bloat in the code did contribute to the early difficulties users experienced. The site design was heavy with Javascript and CSS files, and it was peppered with small coding errors that became particularly troublesome when users faced bottlenecks in traffic. Frequent typos throughout the code proved to be an additional embarrassment and were another indication of a troubled development process.

NoSQL Database
The NoSQL database is intended to allow for scalability and flexibility in the architecture of projects that will use it. This made NoSQL a logical choice for the health insurance exchange website. The newness of the technology, however, means personnel with expertise can be elusive. Database-related missteps were more likely the result of a lack of experienced administrators than with the technology itself. The choice of the NoSQL database was thus another complication in the development, but did not itself cause the failures.

Another factor of consequence is that the website was built with both agile and waterfall methodology elements. With agile methods for the front end and the waterfall methodology for the back end, streamlining was naturally going to suffer further difficulties. The disparate contractors, varied methods of software development, and an unrealistically short project time line all contributed to the coding failures of the website.

I will begin our blog on Java Tutorial with an incredibly important aspect of java development:  memory management.  The importance of this topic should not be minimized as an application's performance and footprint size are at stake.

From the outset, the Java Virtual Machine (JVM) manages memory via a mechanism known as Garbage Collection (GC).  The Garbage collector

• Manages the heap memory.   All obects are stored on the heap; therefore, all objects are managed.  The keyword, new, allocates the requisite memory to instantiate an object and places the newly allocated memory on the heap.  This object is marked as live until it is no longer being reference.
• Deallocates or reclaims those objects that are no longer being referened. 
• Traditionally, employs a Mark and Sweep algorithm.  In the mark phase, the collector identifies which objects are still alive.  The sweep phase identifies objects that are no longer alive.
• Deallocates the memory of objects that are not marked as live.
• Is automatically run by the JVM and not explicitely called by the Java developer.  Unlike languages such as C++, the Java developer has no explict control over memory management.
• Does not manage the stack.  Local primitive types and local object references are not managed by the GC.

So if the Java developer has no control over memory management, why even worry about the GC?  It turns out that memory management is an integral part of an application's performance, all things being equal.  The more memory that is required for the application to run, the greater the likelihood that computational efficiency suffers. To that end, the developer has to take into account the amount of memory being allocated when writing code.  This translates into the amount of heap memory being consumed.

Memory is split into two types:  stack and heap.  Stack memory is memory set aside for a thread of execution e.g. a function.  When a function is called, a block of memory is reserved for those variables local to the function, provided that they are either a type of Java primitive or an object reference.  Upon runtime completion of the function call, the reserved memory block is now available for the next thread of execution.  Heap memory, on the otherhand, is dynamically allocated.  That is, there is no set pattern for allocating or deallocating this memory.  Therefore, keeping track or managing this type of memory is a complicated process. In Java, such memory is allocated when instantiating an object:

String s = new String();  // new operator being employed
String m = "A String";    /* object instantiated by the JVM and then being set to a value.  The JVM
calls the new operator */

Let’s face it, fad or not, companies are starting to ask themselves how they could possibly use machine learning and AI technologies in their organization. Many are being lured by the promise of profits by discovering winning patterns with algorithms that will enable solid predictions… The reality is that most technology and business professionals do not have sufficient understanding of how machine learning works and where it can be applied.  For a lot of firms, the focus still tends to be on small-scale changes instead of focusing on what really matters…tackling their approach to machine learning.

In the recent Wall Street Journal article, Machine Learning at Scale Remains Elusive for Many Firms, Steven Norton captures interesting comments from the industry’s data science experts. In the article, he quotes panelists from the MIT Digital Economy Conference in NYC, on businesses current practices with AI and machine learning. All agree on the fact that, for all the talk of Machine Learning and AI’s potential in the enterprise, many firms aren’t yet equipped to take advantage of it fully.

Panelist,  Michael Chui, partner at McKinsey Global Institute states that “If a company just mechanically says OK, I’ll automate this little activity here and this little activity there, rather than re-thinking the entire process and how it can be enabled by technology, they usually get very little value out of it. “Few companies have deployed these technologies in a core business process or at scale.”

Panelist, Hilary Mason, general manager at Cloudera Inc., had this to say, “With very few exceptions, every company we work with wants to start with a cost-savings application of automation.” “Most organizations are not set up to do this well.”

I suspect that many of you are familiar with the term "hard coding a value" whereby the age of an individual or their location is written into the condition (or action) of a business rule (in this case) as shown below:

if customer.age > 21 and customer.city == 'denver'

then ...

Such coding practices are perfectly expectable provided that the conditional values, age and city, never change. They become entirely unacceptable if a need for different values could be anticipated. A classic example of where this practice occurred that caused considerable heartache in the IT industry was the Y2K issue where dates were updated using only the last 2 digits of a four digit number because the first 2 digits were hard-coded to 19 i.e. 1998, 1999. All was well provided that the date did not advance to a time beyond the 1900’s since no one could be certain of what would happen when the millennia arrived (2000). A considerably amount of work (albeit boring) and money, approximately $200 billion, went into revising systems by way of software rewrites and computer chip replacements in order to thwart any detrimental outcomes. It is obvious how a simple change or an assumption can have sweeping consequences.

You may wonder what Y2K has to do with Business Rule Management Systems (BRMS). Well, what if we considered rules themselves to be hard-coded. If we were to write 100s of rules in Java, .NET or whatever language that only worked for a given scenario or assumption, would that not constitute hard-coded logic? By hard-coded, we obviously mean compiled. For example, if a credit card company has a variety of bonus campaigns, each with their own unique list of rules that may change within a week’s time, what would be the most effective way of writing software to deal with these responsibilities?