Saturday, September 16, 2006

Quality Index (QI): Measure of Risk

Is it possible to capture the quality of an offering in a single metric, something like a Quality Index?

Before I answer this question, I'd like to reflect back on the definition of quality:

Quality: The totality of features and characteristics of a product or service that bear on its ability to satisfy stated or implied needs. (ISO 8402: 1986, 3.1)
Note that the definition refers to stated and implied needs. What that means is Quality of a product which is considered "high" today (because it totally satisfies stated and implied needs) can go down tomorrow because of changes in implied needs. How do we measure that? We probably can't!

Let's re-write the questoin then,

Is it possible to capture the intrinsic quality of an offering in a single metric, something like a Quality Index?

This sounds more reasonable. But, there are over 100 metrics that can be argued to impact the quality of an offering. How do we make sure that we are capturing everything and our QI is based on the perfect algorithm?

Good news is that we don't need to look into hundreds of metrics. 80/20 rule applies here too! We can take 20% of the top variables to get the 80% of insight into intrinsic quality.

What if there is an error in the algorithm we choose? It is possible that our QI is off by 10% or even 20%. If e represents the error, then QI(Observed) = QI (Real) +/- e

If we use the consistent mechanism to capture QI, same error (e) will exist everytime we take a snapshot. And the best part is that majority of this error will cancel out when we plot QI numbers to look at its trend over time. The whole graph will be offset by e.

Note: The key to quality is consistency and repeatability. It is not really important what process we follow, as long as we can make sure that we can make it repeatable. Same concept applies to QI.

So, the answer is YES. We can capture the intrinsic quality of an offering in a single metric and use the trend analysis to keep tab on improvement. Remember TQM, it is all about continuous measurement and continuous improvement!!

Why calculate QI?

QI is not just a measure of quality, it is also a measure of risk associated with quality for your product offering. Here are some of the real advantages of having a QI for all your projects:
  • Get an insight into release readiness, especially in the agile development
  • QI dependency matrix can help identify problems much quicker in the SOA world
  • Mapping QI ranges with your Customer escalation data can give you an insight into what to expect when you release a product with QI less than 60 as compared to a product with QI greater than 90.
  • QI trend provides continuous feedback - required for control. It is easy to monitor when process is going out-of-control.
  • Easy for management to digest one number and drill down, if required.
Remember:
  • QI like number in itself is probably not meaningful. It's the trend that is relevant.
  • QI number can't be compared across companies in the industry, as there are a lot of variables in the "error" part, which won't cancel each other!
  • QI probably can't even be compared across different teams with-in the same company
Further Reading:

Tuesday, September 05, 2006

Only 7% say SOA exceeded expectations!!

Interesting survey on SOA Adoption: "The Dark Side of SOA"

Out of 273 business and technical professionals:

  1. 24% say projects fall short of expectations
  2. 55% say SOA introduced more IT complexity
  3. 41% say SOA projects cost more than expected
  4. Only 7% say SOA exceeded expectations!!
According to HP's SOA executive director, Terri Schoenrock, these failures are because of "misalignment between business and IT goals." That's probably true. To add to this, I think, the program management teams are the ones that don't understand the power of SOA and how it impacts different functional groups. Since PMs are the facilitators and co-ordinators, the failure of SOA projects fall on their shoulders.

The article also mentions "As software grows over hundreds and thousands of components slated for reuse, and Web Services extend beyond company walls, having right tools to manage them is critical."

Probably it's not even management's fault. It is the lack of right tools that has crippled their ability manage SOA projects and their interdependencies effectively. SOA is about busines agility and continuous change!

By reducing the size of projects, and increasing the overall number of projects, SOA has created problems for the IT organization and the management. Companies don't have right tools that can help project managers to manage more projects, QA to test more efficiently and IT to deploy more frequently. Legacy tools are tied with legacy processes, which doesn't fit perfectly for SOA needs.

New technology (SOA) is pushing for new processes (in IT, Development, Project Management, and QA) and the new processes won't be efficient without the new tools.

The article also points at the SOA governance issues for the failures. According to IBM's VP of Web Services and SOA, Michael Liebow, companies with savvy executive teams tend to be most efective at governing SOA projects!

Tracback URL: Only 7% say SOA exceeded expectations!!

Saturday, September 02, 2006

Understanding SOA

SOA is the buzz word these days. Gartner reports predict that more than 80% of applications will be based on SOA by year 2009. So, what is SOA and why is it picking up so fast?

SOA means different things to different people. Different parts of organizations have different motives behind adopting SOA. It has something for everyone!

We all know SOA is "Service Oriented Architecture", but what does it really mean? And more importantly what does it mean to an engineer? What does it mean to an IT organization? What does it mean to the top management? And what does it mean to a company as a whole?

Distributed Computing

Distributed computing is not new. Over the past 2 decades, we have seen it grown from simple 2-tier client-server architecture to 3-tier and then N-tier and p-2-p architectures. The goal of a distributed system has always been scalability, availability, and flexibility. Moreover, distributed architectures have helped companies to better align their resources (both human and hardware)

SOA evolution

Distributed computing allowed companies to break their systems into smaller components. This helped companies to realize the fact that a majority of these smaller components are not related to their core business and, in fact, are common across different companies and even industries. For example, entitlement and billing systems, payment systems, catalogue management, application infrastructure layers and even middle wares.

With the bust of dot-com era, companies started thinking about ways to lower costs. One of the most obvious was to out-source the non-core aspects of the business. The outsourcing and consulting companies find this really interesting, as they started leveraging their experience across companies and industries. The need for openness and ability to consume third party services led to the advent of web-services and Service Oriented Architecture (SOA).

Software as a Service (SaaS) and SOA

Software delivery as a service has grown rapidly as more and more companies have realized that they can keep the customer relationship active by delivering software as a service, rather than delivering one-time products. I see SOA as a SaaS enabler.

What does SOA mean to an Architect?

SOA provides loosely coupled components. Each component is implemented using open standards for flexibility and future expandability. Architects need to make sure that there is no Single Point of Failure (SPOF) in the end-to-end system and, in some cases, are forced to implement extra caching mechanisms to improve availability and reliability. SOA brings simplicity back to architectures.

What does SOA mean to Development?

To developers, SOA means smaller teams and shorter development cycles. However, SOA also means more coordination across different teams and requires developers to follow standards.

What does SOA mean to an IT organization?


SOA enables faster release trains. With SOA, not all components needs to be upgraded at once. That means Bug fixes can be delivered faster, safer and easier. Less downtime and happy customer. SOA provides business agility. SOA enables continuous change!

What does SOA mean to Management?

SOA means large number of smaller projects rather than one monolithic project. Till date, it has been a nightmare for management, as they have not been able to grasp the power of SOA and the concept of smaller projects. To management, it only means more projects. Since Software development has been famous for not following discipline and processes, SOA has caused more harm than good. SOA requires engineering discipline and standardization.

A lot of organizations have been caught in the messy middle - the market environmental pressures are forcing them to enable SaaS and adopt faster release trains, Technology office and architects are pushing SOA, and the managers are stuck with their dogmatic viewpoints around waterfall methodologies. Engineers are confused, not knowing what to do, and quality is the last thing on anyone's mind.

The irony is that SOA goes hand-in-hand with agile development methodologies like Scrum. If and only if the management can be trained in SOA and newer development mythologies like Scrum, the companies can dream of coming out of this messy middle.

Changing market environment and industry pressures have forced companies to adopt SOA, but not everyone in management is even aware of the change. Management needs to be trained with new ALM tools to be successful.

What does SOA mean to a Company?

SOA enabled applications live longer and the ROI curve becomes very lucrative very fast. For companies, SOA is a blessing (if only they can deliver it right).

Final Thought:
SOA requires better ALM tools and integration of processes. The dependency among different components of SOA is not trivial. Sophisticated tools are required to orchestrate the whole application life-cycle management.

Trackback URL: Understanding SOA

Automation: Beyond Testing

Automation is the key to productivity in any functional group in any organization. But for majority of people:

  • "Automation" = "Test Automation"
The general perception is that we should spend all our resources in automating test cases to cut-down test-cycle times! Well, that's not entirely true. Test cases are a small piece of the overall puzzle. Automating test cases is important, but it alone does not gaurantee reduced cycle time and faster release trains.

We must automate not only the overall testing process (which is far more than just automating test cases), but also other processes that takes up considerable amount of resources and are prone to errors. The biggest problem with automating processes is getting buy-in from the top management and all the cross-functional teams they impact. Moreover, some people see automation as a threat to their job and hence never fully support the idea. However, the fact remains that to boost productivity, one must maximize the use of labor and capital - which can be best achieved via Automation!!

Automation has following benefits:
  • An automated process tends to live longer and ROI follow an exponencial curve.
  • Automated processes are a lot easy to follow.
  • Automation provide repeatability and consistency and boosts the quality of the process itself.
  • Good quality processes results in good quality products (TQM principle)
  • Employees get more frustrated by the process relaeted issus. A lot of product related issues are there because of broken processes.
  • Process Automation makes the upper managaemet's life easy. It generates data and trends which can significantly impact decision making.
  • Automated process reduces the ramp-up time for new comers, as processes are a lot easier to understand, once automated.
  • Process Automation is an investment in the infrastructure.
So, how imporant is it for a team to invest in automation beyond testing? The answer is "It Depends." It depends on the complexity of the team, its structure, size and the amount of decentralization. Processes becomes overhead when your teams are small and localized, like in a startup. Most startups hate processes. Whereas the processes become really important when team grows in size, members work from home and are distributed globally in different timezones. Processes also makes sense when team members are of mixed experience as a result of attrition and growth, and there is a growing need for cross-functional collaboration.

In case you are wondering which process to automate, here is a small list of ALM sub-processes, which are a good pick for automation:
  • Requirements Management
  • Version Control, Build and Release Management
  • Change Management, Request to Integrate (RTI) Process
  • Patch Management
  • Request for Change (RFC) Process
  • QA Acceptance, Sanity testing
  • Overall Testing Process (includes test management and test automation)
  • Continuous Integration
  • Deployment
  • Defect Management
  • Customer Escalation Management, Issue Tracking
  • Program Management
  • Monitoring (servers, errors, thresholds, CPU, memory, I/O, response times, availability)
  • End-to-end real-time reporting
  • Dependency Management (must for SOA applications)
Process automation doesn't mean going out of limits and defining a process and then automating it. Start by identidying iterative activites that happen very frequently and are annoying. For large and distributed teams, where it's hard to get a handle on productivity, processes are a breeze. And if there are processes, they need to be automated to bring standardization, acceptance, speed, and higher overall productivity.

Final Thought:
If there is a process, it better be automated!