Reacquainting myself with Java web application development

After a few years in a technical management role, I am reacquainting myself with hands-on Java web application development. After all, nothing says "can do" like "have done" -- however there's an almost bewildering amount of choice when it comes to picking frameworks.

I decided I would start by looking at Spring, Tapestry, Wicket and maybe plain EE6 as that would cover the most popular setups people seem to be using. Additionally I have chosen to use NetBeans as my IDE (I always had a soft spot for NetBeans, and it's looking pretty good these days), Maven as my build tool and JPA/Hibernate for persistence.

I decided to build a web app for managing information about job opportunities. Something that might help keep one on top of things amid the phone calls from recruiters, talent acquisition managers and the various phone screens and interviews.

Minimally, this app would have a way to enter information about jobs, and review the list of jobs. This would all be on a single page. Beyond that there are a few other things I may add once I have the basics in the various different frameworks figured out including:

• Making the form submit / addition of a job AJAX based to eliminate page refresh 
• Editing job information 
• Changing status of jobs (applied, phone screen, in person interview etc.) 
• Bringing in jQuery and looking at some sexy transitions to reveal and hide the "add job" form, sortable table etc. 
• Paging of the jobs table (granted I hope nobody needs to apply for so many jobs that paging is required…but interested to see how this pans out) 
• An entirely gratuitous web service of some kind…

OK, so first up is Spring. A blog post on my experiences there will follow shortly.

Explaining different types of testing

I was recently involved in a software release where some rather unfortunate defects made their way to production. This resulted in some (not unreasonable) tough questioning from our stakeholders.

One particular area they wanted to ask about was our testing practices. Were we not testing our product the same way their user community used it?

This highlighted an important disconnect in understanding the variety of types of testing that a responsible, professional team would usually employ to thoroughly test a product.

It was true that our testing wasn’t good enough for that release. But there was much more to it than simply “testing the software in the way users use it.”

In a bid to help them understand our weak areas (and what we were doing to remedy those) I went hunting for one of those “test pyramids” modeled on the same idea as the USDA food guide pyramid (http://www.nal.usda.gov/fnic/Fpyr/pmap.htm)

There are plenty out there, but I wanted to assemble my own which is what you can see here (click to enlarge):

If I were to do this again I’d probably separate out a bit more the different types of testing in the lower blue section for better emphasis of each, but you can get the gist of things from it as is.

As you can see, I’ve classified the testing appropriate for our product in quite a few different ways. Understanding what these all were, how they differed and why each different kind was necessary was (it became apparent) difficult for our non-technical audience to easily understand.

Reflecting upon this later, I came up with the following analogies that I hope helps make these different kinds of test more understandable to non-developers (and even some developers...)

Why so many different kinds of test?

This too was a puzzle for some. The idea I came  up with to explain this was that, just as a doctor may need a wide variety of tests to diagnose a patient, so too we need a variety of tests to check the "health" of our software.

The analogies below are all based around the idea of building cars. They might be a bet stretched, but I hope they're useful.

Unit tests

Inspecting each component in isolation. Confirm the building blocks are fit for use in the bigger system (car).

Feature/acceptance tests

Checking all features behave as expected, e.g. does it steer properly, doors and trunk open properly, seats adjust etc. 

Performance tests

Driving around a racetrack, time trials, checking how fast you can get from 0-60?

Stability and reliability tests

Driving the car non-stop until you've done 200,000 miles+ without (too many/too severe) problems

System tests

Driving around the neighborhood, highways, driving with passengers, kids etc. (i.e. using it the way real people would do)

User acceptance testing (UAT)

Getting the public to try it.

Exploratory testing

Driving around town, down the highways, mountain roads, 4WD roads, dirt roads, etc. You hope to find unexpected things.

Smoke testing

Giving each car that comes off the production line a quick drive.

If you could point someone to just one link...

...as a quick summary of what agile is, which one would it be?

This is a question I just got asked. And I'm honestly not sure. It would depend a bit on who that "someone" was. But here's a very general definition for agile software development (ignoring the application of agile to other disciplines such as marketing) that I think might be OK.

Agile is an approach to developing software that:

• Focuses on prioritizing the most valuable features first
• Organizes work into short (typically 2 – 4 week) chunks of time typically known as iterations
• Builds and tests one or more complete new features inside each iteration
• Shows and discusses what they accomplished with the customer at the end of each iteration allowing regular feedback and guiding of the effort
• Acknowledges that it is nearly impossible to figure out completely up front what people want from a software system, and that their opinions and business needs will change as they start to see it built. 
• Relies on cross functional teams (developers, testers, whatever else is required)working as a self-organizing group to accomplish their goals. A good example of a self-organizing team outside software is a sports team, e.g. think how soccer players collaborate with the end objective of scoring goal(s).
• Employs a number of technical disciplines to allow incremental development, with a particular emphasis on the importance of regularly integrating and testing code with as high a degree of automation as is sensible

I'm sure there are much better quick'n'simple overviews along the lines of "What is agile?", but it's surprising what you get if you google that. Based on the couple top links I visited it wasn't ideal as an explanation for a  general audience.

Cloudy With Metrics

I’m working on a team that is responsible for a Windows desktop application that allows the display and manipulation of medical images. Medical images are just what you would imagine: x-rays, ultrasound, MRIs, computed tomography etc.

These images are large – hundreds of megabytes (more for ultrasound video) – and therein lies one of our big challenges: high performance for users everywhere. And many of our users are sat in Southern India or Europe, while our images are in a datacenter on the East Coast of America.

Now while we do have the luxury of nice big pipes connecting these locations, they inevitably suffer from fairly high latency (~300ms for India) due to the physical distance between them. That level of latency and TCP do not mix well.

This leads to a very disappointing rate of throughput even when ample bandwidth is available. Improving this situation can be boiled down to the application of just three techniques:

1. Move the data faster
2. Move less data
3. Move the data before you need it

·        For us, the option of moving the data before it was needed was already off the table. Pre-caching image data is an approach we had been using for years with a legacy product. Although it worked respectably well, it was far from transparent to end users. With some regularity, data that people expected to reach remote users had not done so in a timely fashion, and much gnashing of teeth and submitting of helpdesk tickets ensued. Caching was a four letter word, so to speak.

So, we focused on moving the data faster, and moving less of it to boot. Exactly how we did that is not really the focus of this post (but so as not to leave you hanging, scaling of images and UDP were key).

What this post is really about, is how we made use of Amazon’scloud-based SimpleDB as a super-simple means to capture in-the-field metrics on the performance and usage of our application.

In our earliest pilot phase we displayed time to load and throughput in the application’s status bar, and users could obviously relay that information to us (yeah, that was as unsatisfactory as it sounds). And of course it was equally possible to log things to a user’s machine, e.g. with Log4Net or using the Windows Event Log. But that approach makes access to, and consolidation of, the captured data a bit more involved than we wanted it to be. So although it’s not that we couldn’t capture this data without a cloud based solution, it’s clear that some single, centralized capture of this data was an appealing option.

One way we might have got a centralized means to capture this data would be to have built ourselves a small database, fronted it with a set of web services and used that. But the effort involved in getting that done and deployed is, as with many organizations, non-trivial and we were looking to slip this functionality in ASAP.

By comparison, Amazon’s SimpleDB is exactly that – simple.

It’s “schema-less”, so there’s no DDL to create tables and so forth. Instead, data is stored in domains (akin to tables; you do need to create these ahead of time) made up of items (just like a row) as a set of attributes (a bit like a column, but better thought of as name-value pairs).

But enough English, here’s some C# to show just how easy it can be to stash some data into SimpleDB. It can in fact be even easier than this, since here I’m using the asynchronous approach to putting data in.

       private void LogMetricToCloud(IMetric metric)

        {

            string domain = metric.GetType().FullName;

            CreateDomainIfNeeded(domain);

            try

            {              

                PutAttributesRequest request = new PutAttributesRequest()

                    .WithDomainName(domain)

                    .WithItemName(Guid.NewGuid().ToString());

                foreach(var attribute in metric.Attributes())

                {

                    request.WithAttribute(new ReplaceableAttribute()

                        .WithName(attribute.Name)

                        .WithValue(attribute.Value));

                }

                string state = metric.GetType().FullName + ":" + metric.CSVData();

                IAsyncResult asyncResult = _db.BeginPutAttributes(request, new

AsyncCallback(SimpleDBCallBack), state);

            }

            catch (Exception e)

            {

                TryLogError(e.Message);

            }

        }

        private void SimpleDBCallBack(IAsyncResult result)

        {

            string state = result.AsyncState as string;

            try

            {

                // If there was an error during the put attributes operation it will be

thrown as part of the EndPutAttributes method.

                _db.EndPutAttributes(result);

            }

            catch (Exception e)

            {

                TryLogError(string.Format("Exception: {0}\nFailed to log: {1}", e.Message,

state));

            }

        }

I think that what the code does and how is fairly self-evident, but that could just be because I’ve been working with it. Just to make sure it’s clear, here’s a breakdown:

• I have a number of different classes of metric, all conforming to the IMetric interface
• I have a domain (think: table) for each different kind of metric
• As a concrete example, one of my metrics measures feature utilization; features are things like flipping and masking images. For feature utilization I record who, where, when, what project and what feature they used.
• The  LogMetricToCloud method simply iterates through all the attributes of a metric and creates a request to persist this in my SimpleDB instance.
• The request is made asynchronously, with the creatively named  SimpleDBCallBack method being invoked upon completion.  The call here to EndPutAttributes will let us know if any problems occurred with the original request

All of this seems to be working pretty well, and I hope that in the near future we can start to make even more use of SimpleDB. Next on the list for me is implementing some “feature toggle” type functionality for our application. With that we can entertain some of the ideas described in this great article by Jez Humble about core practices for continuous delivery.

DICOM non-technical introduction: mind map

Earlier today I went through the non-technical introduction to DICOM on the RSNA (Radiological Society of North America) website written by Steven C Horii, MD. As I was doing so I compiled a mind map using the excellent SimpleMind tool to help jog my memory on this stuff in the future.

Click the image below for the full size PNG or it's also available here as a PDF.

Understanding progress: on points, velocity and when to add new stories

Building software takes time. Usually enough time that people are interested in monitoring progress and understanding when it will be done. Agile teams often use User Stories as a unit of work. Typically these are estimated in points enabling a team to record their velocity, that is, the number of points completed per sprint.

With this data, teams have a simple means to show progress. Interested parties can follow along quite easily, seeing how each sprint eats away at the features in the backlog and how many points are left to complete the product. Using the team’s average velocity will give a nice indication of how many more sprints are required to complete the features in the backlog: points remaining ÷ average velocity = sprints remaining.

There is one wrinkle to this otherwise simple scheme. As anyone who has developed software can tell you, the devil is in the detail. Something that initially looks simple can end up being more involved than originally estimated. Case in point: my team recently had a story like the following for a desktop client image processing product they are building:

 
     As an imaging assistant
     I want to be able to open JPG images
     So that I can process them

The story seemed to be straightforward and was implemented easily and quickly. All was well until we tried some particularly large JPG files. At this point things blew up with out of memory errors and the like.

So was handling large JPGs a new feature and thus a new story? Or was it obviously part and parcel of the original? In other words, the question is how do you deal with this and still report easily understandable progress?

The short answer is it doesn’t really matter. You can add new stories to the product to cover the new work that you discover. Or you can just do the work that was implied but not necessarily obvious from the original story. The formula for figuring out how many sprints remain still works either way. Taking the former approach your velocity is likely to remain fairly stable. Taking the latter approach you may see it bounce around a little bit more, or see it dip from historic levels if the team had a velocity established through more predictable types of work (e.g. maintenance on a well known product.)

The longer answer is that each approach has its own pros and cons. Depending on your situation one may be better than the other.

Adding a story, pros:

• Stakeholders can see the amount of work we original imagined was involved has grown and have more chance to comment on the necessity of these items – perhaps the complexities and edge cases discovered aren’t that valuable.
• It probably shouldn't matter, but velocity remains stable and nobody feels the need to cross-examine the team and ask "Why has your velocity dropped?" Done this way velocity may even serve as a crude indicator of team performance and improvements can be seen in higher velocity.

Adding a story, cons:

• It probably shouldn't matter, but there's likely a class of stakeholder that will question why all this "extra" work is emerging and ask how come we failed to identify it in the first place.
• If the team has the pleasure of using a computerized agile lifecycle management tool then each additional story is another thing to enter, estimate, prioritize, track and update status on etc.
• The number of points to complete the originally envisaged release keeps growing making some people anxious: "We don't know how much work is left to do, how can you ever hope to predict when it will be done?"

Not adding a story, pros:

• Nothing extra to track (though we probably need to clarify the acceptance criteria of the original story)
• Number of points to complete original feature set for the release remains the same (unless we discover a need for genuinely new features)

Not adding a story, cons:

• The potential roasting of the team: "Why is your velocity erratic/dropping?"
• The team might miss an opportunity to push low value work down the backlog.

Personally I'm most strongly drawn to the idea of not adding in extra stories. I like the simplicity and minimalism of this. The more items there are in the backlog the harder it is to grok the thing as a whole and more busywork goes into managing it all. I don’t think the “pros” of adding in extra stories are powerful enough to make it the preferred approach. And although there is the potential “con” of the team getting questioned about why their velocity is erratic or dropping, I believe this can be explained quickly in simple terms. I also think it’s a lot more straightforward, even comforting, for stakeholders to see that the size of the backlog remains fairly stable unless things they too understand as new work (features) get added.

Enemies of Agility: The Dirty Dozen

Developing software is hard. Agile software development represents one way to deal with the complexity and difficulty in a manageable fashion. Arguably the best way we've figured out so far. But it's not easy, and there are many impediments on the road to becoming a highly adept agile team. Below I present the "dirty dozen" impediments, or key enemies of successful agility that I've observed.

1. Management doesn’t understand agile
An individual team or two can get quite a lot of improvement by themselves. But to really scale up agility in an organization requires a deep understanding and commitment to the guiding principles from managers across a range of functional areas. Attaining that in a large corporate environment is surely hard. Even among the engineering or IT groups that might have more capacity or interest to “get” agile at a management level there are challenges. How can people many years distant from the act of actually creating software really understand what changes are occurring on individual delivery teams if they haven’t experienced them first hand themselves? How can they help shape and encourage greater adoption and performance, policies and culture if they are not actively engaged in a deep meaningful way with agile? How can they negotiate effective cross-department agreements when they don’t really understand how their teams are working?

It’s hard. Not impossible though. What it requires is a willingness at the management levels above individual delivery teams to immerse themselves in as much agile as they can. Go to conferences. Get the deep dive that your two day agile orientation course didn’t give you. Read books. Lots of books. Join mailing lists and discussion groups. Read the blog posts people put out there. Follow the thought leaders on Twitter. Participate on a team for an iteration – there lots of ways for people of different skills to do this. Pair with the PO to groom the backlog. Pair with a coder and/or tester. Do some exploratory testing, etc. etc.

Once managers have experienced being and doing agile they have a much greater chance of extending those principles up and across the organization.

2. Team doesn’t understand agile
Understanding agile requires more than sending one (or more) of the team off for a two day Certified Scrum Master course. And it’s considerably more than saying “go do agile.” I’ve talked with lots of people for whom this is all they’ve had, and it’s clear that the depth of understanding is far shallower than it needs to be to really improve.

From my experience two things are required for a team to really understand agile deeply rather than just go through the motions. Firstly, as suggested above with managers, people on agile teams need to really immerse themselves in the wider agile world. There is an immense volume of great stuff to discover, read, try and learn out there. Harnessing that will help a team understand what other people are doing and being successful with. Secondly, the team needs a coach, at least at the outset. The coach needn’t be an externally hired consultant; it could simply be an experienced scrum master or effective manager. Someone who can help guide the team away from common pitfalls and remind them of common pitfalls.

3. Team doesn’t understand end users
How can you make intelligent decisions about the software you are building and testing unless you have a deep understanding of the way it is used? Not understanding the product can lead to several problems: PO or some other key figure becomes a bottleneck; requirements become gospel and no intelligent interpretation is allowed; stupid things get done because people misunderstand needs; etc.

Depending on who and where your users are vis-à-vis the team this may or may not be easy to address. Ideally your users are accessible and team members can spend time with them regularly to understand better the world in which they work. Where this isn’t possible some effort should be put into finding ways to make sure the team understands the users as much as possible.

4. (Micro)managers who’re not on the scrum team interfering
This can be a particularly debilitating and frustrating class of dysfunction. How is a team meant to develop the ability to self-organize and self-manage if team member’s managers regularly disrupt matters?

This can manifest in a variety of ways:

• Dictating how work should be done
• Dictating priority thus undermining the PO
• “Refining” or challenging estimates
• Focusing on velocity over quality and predictability
• Injecting additional work
• Making promises to external parties without consulting the team
• Having conversations that influence technical decisions absent the team
• Confusing stakeholders by offering an uniformed, contradictory and out of date view of status

Managers may indeed have a lot of experience to offer to their teams, but micromanaging interference will sap teams’ engagement. Managers need to realize that they should go about things in a less directive fashion than they may have previously. Rather than push down decisions and work, managers need to harness the talents of a team and present the results up and out to the rest of the organization. Teams typically have (or can develop) regular and agreed upon venues and approaches for managing the how and what and when of their work. Managers who wish to influence things must utilize these to obtain team buy-in and agreement. Failing to do so invites resentment, apathy and underperformance.

5. Inappropriate or suboptimal tools
This is a problem that occurs more in larger organizations where well meaning people want to “standardize” on things. Things like last decade’s monolithic proprietary testing frameworks or agile lifecycle management tools that get slickly pitched to executives offering alluring reports to analyze all their staff but offer limited useful functionality to the actual teams using them day in and day out.

I’m not sure there are any good immediate ways to address this kind of problem. Perhaps when management obtains a deep understanding of agile principles such issues will diminish. In the interim, if the subversive option of simply going your own way, irrespective of corporate mandates is a possibility, I say go for it.

6. Poor technical and problem solving skills
Building anything but the most trivial software is hard. It requires good people. People with excellent technical and problem solving skills. This is as true of traditional approaches to software development as it is to agile of course. Good people will do better with a bad process than bad people with a good process.

Agile software development tends to “out” people who don’t make the grade more so than traditional forms of development where people could go weeks or months without having too much scrutiny applied to their work. The focus on building small, incremental, potentially shippable features in short blocks of time especially needs people with good technical and problem solving skills.

People need to be able to write good code: well designed, well factored, elegant and simple. They need to understand how to effectively use unit tests, to understand the benefits of continuous integration, pair programming, TDD, source code control, automated testing and on and on. They need to be able to approach problems with an engineering mindset, to solve problems scientifically, methodically. A team absent these skills will have problems.

7. Poor written and verbal communications
Having the technical and problem solving skills mentioned above is not enough. People need to be able to communicate effectively too, especially on agile teams.

When people are unable to explain themselves coherently there is a massive drag on the team (and potentially outside the team too). People have to clarify what people are saying are writing about. People have to dig for more information that is absent or not understandable in the original communication. People have to guess and may misunderstand. All of this is frustrating and costly.

8. Apathy
“It’s not that I’m lazy, Bob. It’s that I just don’t care.”

Some people are bored. Some people don’t understand what the purpose of their work is. Some people just want to be left alone to hack code and they perceive that agile gets in the way of that. Some people just want to be left alone to cruise Facebook. Whatever the cause, apathy will drag a team down.

Managers have a large stake in combating apathy. Assuming your company is doing something purposeful make sure people understand what it is, why it’s important and how they can help. Make sure they feel empowered to influence the work, the team, the tools. Help create intellectually interesting challenges that marry business needs with opportunities to grow. Support their learning and training needs. Shield them from soul sapping pointless drudgery. In short, make them give a shit.

For those wondering: what if we grow all these people with training, give them an opportunity to develop new skills and master new technologies and then they leave? Consider this: what if we don’t and they stay? Unmotivated, low-skilled, bored people will not a successful product make.

9. Inability to create good user stories
Admittedly one can quite happily apply agile software development principles without user stories. However, they currently sit as the most popular unit of work for agile software development.

A common “starting package” for a new agile team is comprised the set of Scrum ideas plus user stories as your unit of work, a product backlog and estimation using story points. For a team to get familiar with this is fairly easy. Scrum proscribes a set of standard meetings: sprint planning, daily scrum, sprint review and retrospective. Estimating in points requires a little more explanation and practice typically but eventually seems to work nicely for most.

The hardest part though in my experience is to understand and employ user stories well. The five most common problems seem to be:

• Stories for everything: setting up new servers, writing a summary of findings, having a meeting, etc. etc. Stories are about features the software should have that make sense to users of the product. This is closely related to the misunderstanding of velocity item below.
• Stories that are tasks: this is similar to the stories for everything problem, except that even when your stories are constrained just to product features it’s still possible to write them as discrete tasks rather than describing a full end-to-end feature.
• Stories artificially bounded by system architecture: It’ll depend on your users, but for most users of most applications they do not know or care that your product is made up of different components. Devising stories for “the server” and “the client” may well be an indication that you’re not thinking about features from an end user point of view but from a system design perspective.
• Abuse of the “as a  I want  so that ” formula: for many people a story is simply stating your requirements in this format. This often leads to rather convoluted looking stories. Couple this with the stories for everything or stories that are tasks problems above and you end up with things like “As the scrum master I want some administrivia completed so that I can report it to my boss.” This is pretty far adrift from where you want to be.
• Poor acceptance criteria: conveying who wants what and why is a good start. But it is also essential to bound the scope of the story with good acceptance criteria. Ambiguity leads to confusion, poor estimates, frustration and potentially disappointment.

There’s no easy cure for this problem. You need someone on the team or someone coaching the team to a point where they have a good understanding of user stories. That or everyone reads Mike Cohn’s User Stories Applied…

10. Lack of automated testing
Whenever you change software there’s always the risk that you unintentionally break something. Sometimes you can make good educated guesses about what may break as a result of your changes. Other times the most innocuous seeming change creates a cascade of fail in the most unlikely of places. To make sure your changes don’t break things that were working before you need to test your software.

If you test it manually then the more your software grows the more time you need to execute your tests. One can imagine how, over time, more and more of an iteration could be taken up with making sure the newly added features haven’t inadvertently broken those that were already there. Developers are left with nothing to do and test staff are run ragged trying to keep up. Clearly a bad situation.

While an automated test suite may take more and more time to run it’s nothing like the days or more that manual testing requires.

You simply cannot hope to incrementally add features, release regularly and maintain quality without automated testing.

11. Misunderstanding the purpose of velocity
Velocity is not a measure of all the work a team has done. It’s a measure of how quickly a team can turn a backlog of product features into working software. When people add stories for training or support or infrastructure maintenance, sometimes retrospectively, ostensibly to “capture the work” or “earn credit for what we’ve done” this pollutes your velocity. It no longer provides an accurate measure of how much a team can advance a product. You might as well record that people work 40 hours a week.

By measuring velocity as just how many product features the team can implement in an iteration, and treating everything else as overhead you are in a much better position to predict how long it might take to burn through the rest of the backlog.

One of the reasons this problem can arise is a simple lack of understanding. That’s easy to correct. The more challenging problem is when teams are being evaluated by velocity: why did your velocity drop during the last iteration? Why isn’t your velocity increasing since we added Bob to your team? That thinking is missing the point. The value of velocity is improving your ability to predict when a product will have a certain set of features completed. It is not a means for measuring productivity.

12. Toxic team member
The toxic team member comes in many guises: antisocial, incompetent, arrogant, slacker, anti-agile skeptic or just general poor team fit. Irrespective of the particular flavor, one bad apple can cause a lot of damage.

I believe everyone deserves a chance to change. Some frank feedback may help. But, ultimately, if someone’s not working out they need to be…redeployed elsewhere.