What are alternatives to the Waterfall model_问答_开发者

Can 开发者_StackOverflow中文版you please give a methodology that stands to alleviate the disadvantages of waterfall model?

The problem with Waterfall is that it consists of monolithic stages, each building on the previous stage. So the code is developed in one chunk after the entire system has been designed, which in turn happened after all the requirements have been gathered and signed off.

This is problem because any change has to be ratified by a complex procedure and rippled through all the stages. But the lesson of history is: change happens. The requirements are always incomplete, or mis-specified or simply out-of-date by the time we get to coding. Too often design and build proceed on the basis of assumptions which are nullified when the system gets to UAT. This leads to frantic re-work and slippages.

The truth is not many customers are good at the sort of abstract thinking required to envisage a working software software system. And too many IT professionals lack the experience necessary to understand business logic. Waterfall refuses to accept these truth.

The only honest requirement specification is "I'll know it when I see it". So it is crucial to get working software in front of real users as soon as possible. Any methodology which focuses on delivering working software incrementally in short iterations will "alleviate the disadvantages of waterfall model".

Originally that was RAD or DSDM. Then XP tok up the banner. Now there is Agile and related things like Scrum and Kanban.

So why do people persist with the Waterfall method?

There is a common perception that Agile is just a cover for cowboy hackers to ditch all the boring process stuff and get on with what they enjoy most: writing code. The branding of "Extreme Programming" certainly encourage this thought, and, let's be honest, it is not an unfounded allegation. That is, some coders pretend to be agile as an excuse not to plan, design or document. This does not reflect the actual practice of Agile, which require just as much rigour as any other methodolgy.

Also Agile requires a much greater commitment of time from the customer's staff, which many organizations are loath to accept. Also the people footing the bill may be unwilling to empower their junior staff to make decisions. There is an important distinction between Customer and User.

When it comes to outsourcing the waterfall model provides an easy framework for matching deliverables to staged payments. Indeed the contractual aspect maybe stronger than that: in the EU Waterfall is mandated for all projects valued at EUR 100m or more.

Finally, there are projects where Waterfall works well. These projects have knowledge domains which are stable and well-understood by both the customers and the developers.

last word

Despite its failings Waterfall has delivered many projects successfully. This is because hard work, aptitude and integrity are more important than methodology.

The waterfall model was documented in 1970 by a Dr Winston Royce in a paper titled 'Managing the development of large Software Systems'. Basically outlining his ideas on sequential development. His idea was that software could be produced in a similar fashion to an automobile, where the vehicle is pieced together in sequential/linear phases.

This linear approach doesn't really allow for changes in a piece of software once it begins. There is no tight relationship with the end user/client so its harder to outline possible problem areas.

Its worth noting some phases of the waterfall model allow for 'splashback' whereby there is enough time in the development period to go back and make small changes. Time constraints and the amount of work involved and budgets don't really allow for much change if any to be made using this model.

The waterfall model is old, as time goes by software paradigms themselves change. Object Oriented programming is popular, back then it was barely alive. Through the use of the waterfall model its obvious that the flaws have been spotted and this has lead to the alternative development methodologies.

Ok, so now for alternatives. Incremental model is described by Alistair Cockburn(2008) as a staging and scheduling strategy in which various parts are developed at different times or rates and integrated upon completion of that specific part.

Basically incremental looks a lot like this:

Analysis->Design->Code->Test
  Analysis->Design->Code->Test
  Analysis->Design->Code->Test

Number of benefits include lifecycle being flexible and allowing for change from the get go. Working software or rather parts are generated quickly and early on. Code produced is earlier to test and manage due to the small iterations of progress. Not all of the requirements of the system are gathered up front, just an outline. This allows for a quick start, however it might be a disadvantage in some systems as things like the system architecture being supported might be missed.

Iterative on the other hand allows parts of the system to be reworked and revised to improve the system. Time is set aside to allow for this. Iterative does not start with a full specification of requirements. Development is done by specifying and implementing just part of the software. Software is reviewed in order to identify further requirements.This is more of a top down approach. Disadvantages with this methodology are making sure all the iterations are compatible. As each new iteration is approved, developers may employ a technique known as backwards engineering, which is a systematic review and check procedure to make sure each new iteration is compatible with previous ones.A major benefit with the constant iterations is that the client is kept in the loop and the final product should meet the requirements.

Iterative approach diagram.

Other methodologies include Prototyping. Evolutionary and Throwaway. These are also deemed as more of a top down approach. Both process are borrowed from engineering.In engineering it is common to construct scale models of objects to be built. Building models allows the engineer to test certain aspects of the design. The software development prototyping methodology provides the same ideology. Prototyping is not seen as a standalone, complete development methodology but rather an approach to handling selected portions of a larger, more traditional development methodology.

Throwaway Prototyping - Throwaway prototyping does not preserve the prototype that has been developed. In throwaway prototyping there is never any intention to convert the prototype into a working system. Instead the prototype is developed quickly to demonstrate some aspect of a system design that is unclear. It can also be developed to help users or clients decide between different features or interface characteristics. Once any problems or uncertainty has been addressed the prototype can be ‘thrown away’ and the principles learned used in the design and documentation of the actual product.

Evolutionary Prototyping - In Evolutionary prototyping you begin by modeling parts of the target system and if the prototyping process is successful you evolve the rest of the system from those parts. One key aspect of this approach is that the prototype becomes the actual production system. This process allows for difficult parts of the system to be modeled successfully in prototypes and dealt with early on in a project.

Other areas to look into will include Agile-> SCRUM, Extreme programming, Paired programming etc.

Tried to keep it short but people write books on this sort of stuff and there is so much to discuss.

Might be worth having a look at: Incremental and Iterative

The alternative to the waterfall method is "doing it the correct way".

Waterfall seems to make sense if you are on a factory floor assembly line. But I've never seen it work as part of the design process...and sofware development is ALL a design process. And so the waterfall method never really works in the sense that it doesn't help facilitate the creation of high quality product, but rather focuses on process. Process can be great, but what's the point if the product it produces is second rate?

Kanban and Scrum are two of the most commonly used alternatives to Waterfall. I tried to give a good overview and comparison of the different SDLC approaches.

Waterfall relies heavily on massive monolithic phases as mentioned by APC. This is a huge weak point because trying to determine the end product from the start is a fruitless endeavor.

Kanban is slightly cowboy, but I find if you couple it with standups it certainly still has it's place.

Scrum is great for putting pressure on the team and getting ownership on tickets. I've found most places have been going with this one but the downfall of it is some people go overboard with having meetings for everything. Sprint planning meetings, sprint kickoff meetings, daily standup meetings that last 1 hour with 20+ people present, demo meetings, and then finally the post-mortem.

Remember that agile is only as good as you make it and you can easily sink any methodology if you go wild with unrestrained meetings which aren't adding value. Keep it as lean as you can without it being chaotic.

From the top of my head, I can think of ways to palliate the shortcomings of the waterfall model:

Have the coder concentrate on automating the process itself. Automate the transitions between one step and another, so that changes will flow more or less automatically.
Make the process more bidirectional. One principal characteristic in the waterfall model is that changes flow from top to bottom. This is a unidirectional process, and that is part of the problem.

Another thing which would help is (as someone mentioned in an earlier answer) is for the developer to gain a better understanding of the business logic involved, and of what the customer wants, and for the customer to gain knowledge about the characteristics of the development process.

Here are some links about Waterfall model:

http://www.cs.odu.edu/~zeil/cs451/Lectures/01overview/process2/process2_htsu2.html

http://www.buzzle.com/editorials/3-13-2005-67039.asp