The Danger of Centralized Workflows

Wednesday, July 13th, 2011.

It isn’t uncommon for me to have a client or student at one of my courses ask me about some kind of workflow tool. This could be Microsoft Workflow Foundation, BizTalk, K2, or some kind of BPEL/orchestration engine. The question usually revolves around using this tool for all workflows in the system as opposed to the SOA-EDA-style publish/subscribe approach I espouse.

The question

The main touted benefit of these workflow-centric architectures is that we don’t have to change the code of the system in order to change its behavior resulting in ultimate flexibility!

Some of you may have already gone down this path and are shaking your heads remembering how your particular road to hell was paved with the exact same good intentions.

Let me explain why these things tend to go horribly wrong.

What’s behind the curtain

It starts with the very nature of workflow – a flow chart, is procedural in nature. First do this, then that, if this, then that, etc. As we’ve experienced first hand in our industry, procedural programming is fine for smaller problems but isn’t powerful enough to handle larger problems. That’s why we’ve come up with object-oriented programming.

I have yet to see an object-oriented workflow drag-and-drop engine. Yes, it works great for simple demo-ware apps. But if you try to through your most complex and volatile business logic at it, it will become a big tangled ball of spaghetti – just like if you were using text rather than pictures to code it.

And that’s one of the fundamental fallacies about these tools – you are still writing code. The fact that it doesn’t look like the rest of your code doesn’t change that fact. Changing the definition of your workflow in the tool IS changing your code.

On productivity

Sometimes people mention how much more productive it would be to use these tools than to write the code “by hand”. Occasionally I hear about an attempt to have “the business” use these tools to change the workflows themselves – without the involvement of developers (”imagine how much faster we could go without those pesky developers!”).

For those of us who have experienced this first-hand, we know that’s all wrong.

If “the business” is changing the workflows without developer involvement, invariably something breaks, and then they don’t know what to do. They haven’t been trained to think the way that developers have – they don’t really know how to debug. So the developers are brought back in anyway and from that point on, the business is once again giving requirements and the devs are the one implementing it.

Now when it comes to developer productivity, I can tell you that the keyboard is at least 10x more productive than the mouse. I can bang out an if statement in code much faster than draggy-dropping a diamond on the canvas, and two other activities for each side of the clause.

On maintainability

Sometimes the visualization of the workflow is presented as being much more maintainable than “regular code”.

When these workflows get to be to big/nested/reused, it ends up looking like the wiring diagram of an Intel chip (or worse). Check out the following diagram taken from the DailyWTF on a customer friendly system:

stateModel

The bigger these get, the less maintainable they are.

Now, some would push back on this saying that a method with 10,000 lines of code in it may be just as bad, if not worse. The thing is that these workflow tools guide developers down a path where it is very likely to end up with big, monolithic, procedural, nested code. When working in real code, we know we need to take responsibility for the cleanliness of our code using object-orientation, patterns, etc and refactoring things when they get too messy.

Here is where I’d bring up the SOA/pub-sub approach as an alternative – there is no longer this idea of a centralized anything. You have small pieces of code, each encapsulating a single business responsibility, working in concert with each other – reacting to each others events.

Productivity take 2: testing and version control

If you’re going to take your most complex and volatile business logic and put it into these workflow tools, have you thought about how your going to test it? How do you know that it works correctly? It tends to be VERY difficult to unit-test these kinds of workflows.

When a developer is implementing a change request, how do they know what other workflows might have been broken? Do they have to manually go through each and every scenario in the system to find out? How’s that for productivity?

Assuming something did break and the developer wants to see a diff – what’s different in the new workflow from the old one, what would that look like? When working with a team, the ability to diff and merge code is at the base of the overall team productivity.

What would happen to your team if you couldn’t diff or merge code anymore?
In this day and age, it should be considered irresponsible to develop without these version control basics.

In closing

There are some cases where these tools might make sense, but those tend to be much more rare than you’d expect (and there are usually better alternatives anyway). Regardless, the architectural analysis should start without the assumption of centralized workflow, database, or centralized anything for that matter.

If someone tries to push one of these tools/architectures on you, don’t walk away – run!

If you liked this article, you might also like articles in these categories:

If you've got a minute, you might enjoy taking a look at some of my best articles.
I've gone through the hundreds of articles I've written over the past 6 years and put together a list of the best ones as ranked by my 5000+ readers.
You won't be disappointed.

If you'd like to get new articles sent to you when they're published, it's easy and free.
Subscribe right here.

Follow me on Twitter @UdiDahan.

Something on your mind? Got a question? I'd be thrilled to hear it.
Leave a comment below or email me, whatever works for you.

51 Comments

Simon Munro Says:
July 13th, 2011 at 3:22 am
Sadly, the same argument can be made against ETL tools (SSIS). Unfortunately ETL visual languages are far more prolific than workflow ones

Pierluigi Says:
July 13th, 2011 at 3:23 am
very true. yet the big challenge in loosely coupled distributed systems is to keep track of the inter-dependencies, so as to know where to do code changes whenever a business rule change.
In my experience we are very bad at that – we still rely too much on PEOPLE remembering the details of the system, rather than a catalog or some automated inventory.

Philip Nelson Says:
July 13th, 2011 at 6:10 am
You know, whether the way your business processes are wired together is done in a bpm tool or not, the complexity of what these processes really are, exists. When you don’t put them in a tool, you tend to focus on just the one event at a time, and that does seem simpler: for the developer. But not for the owner of the processes.

You are absolutely correct about one aspect for most of these tools, they centralize some really basic coding techniques, and a bpm tool is just another notation. I think what will be interesting to watch is if these tools advance as nearly all other notations have that survived, to express more complex branches in a more succinct syntax.

Greger Hagström Says:
July 13th, 2011 at 6:12 am
I must recommend Jim Webber (hope I remember his name correctly).

He has an excellent presentation called Guerilla SOA that pictures this problem perfectly.

Tommaso Caldarola Says:
July 13th, 2011 at 6:26 am
If you have to do integration with legacy system an orchestration engine remains one of better solutions because you have to care your side only, so no issue about unit testing as well as version control.

gigiabbrescia Says:
July 13th, 2011 at 7:11 am
The listed points are true for all visual tools (also for differents purposes – I hated Oracle Report, Crystal Report, Microstrategy and so on) but will be interesting investegation on pro cons on using a centralized orchestration system (on which I can develop without visual tool limitations) vs pub/sub approch, also in case when I have to “talk” with legacy system.

Jimit Ndiaye Says:
July 13th, 2011 at 7:20 am
While I’m a big proponent of the SOA-EDA-style publish/subscribe approach, I don’t think they preclude workflows. I believe the two approaches can coexist. With WF4 for instance, custom activities can be written to interact with a service bus such as NServiceBus, sending commands and listening for events,etc. I’ve found that the visual (bi-directional) communication of intent to business stakeholders provided by workflow technologies can be worth the extra hoops they sometimes force us to jump through. As far the testability story, WF4 does offer a lot more in that respect. While I could probably accomplish the same goals using sagas but for specific scenarios, I like the visual medium that a workflow platform like WF4 provides.

My 2c

David Nelson Says:
July 13th, 2011 at 9:36 am
As usually happens in these discussions, you start out by claiming that centralized workflow is bad, but then try to justify that claim by complaining about the tooling.

“I have yet to see an object-oriented workflow drag-and-drop engine.” – What if you did?

“I can tell you that the keyboard is at least 10x more productive than the mouse.” – Who says a centralized workflow engine has to be operated using the mouse?

“…these workflow tools guide developers down a path where it is very likely to end up with big, monolithic, procedural, nested code.” – There is no reason why a centralized workflow engine couldn’t encourage encapsulation as much as an object-oriented language does.

“It tends to be VERY difficult to unit-test these kinds of workflows.” – But is that because of the nature of the solution, or the inadequacy of the tools?

“When working with a team, the ability to diff and merge code is at the base of the overall team productivity.” – Why couldn’t a workflow tool show you graphically where your diffs are and allow you to merge them?

Integrated development environments, in-depth debuggers, and unit testing suites for mainstream languages didn’t spring up overnight; they have been developed over a long period of time, and they are still developing. That doesn’t mean that the concepts in these languages were invalid until the tools caught up.

If what you are really arguing is that a decentralized approach is better than a centralized approach because the tools to support a centralized approach are not up to snuff yet, that might be a perfectly valid argument. But then you should approach the discussion from that perspective, and the title of your article should be “The Danger of Centralized Workflow Tools”. Then we could have a reasonable discussion about the pros and cons of centralized over decentralized workflow, without dragging in these ancillary concerns.

I for one love the concept of centralized workflow, but I won’t use things like Workflow Foundation specifically because the tools are cumbersome and the results are difficult to debug and maintain. But I look forward to the day when the tooling removes those restrictions.

J.D. Says:
July 13th, 2011 at 11:24 am
Very scary diagram!

udidahan Says:
July 13th, 2011 at 11:46 pm
Pierluigi,

When doing SOA properly, having only one logical publisher for a given event type, there is always only one place where a given piece of functionality/data exists.

udidahan Says:
July 13th, 2011 at 11:47 pm
Simon,

Yes, too much business logic gets thrown in those ETL jobs turning them into their own monstrous broker type environments.

udidahan Says:
July 14th, 2011 at 12:09 am
Phillip,

I don’t think that I’d agree that the complexity of all processes is inherent – many times I’m able to work with clients and simplify their processes greatly by partitioning them. Other times, we see the process give way to a *policy* which is very well modeled using sagas.

udidahan Says:
July 14th, 2011 at 12:10 am
Greger,

Jim and I see very much eye-to-eye on these matters. He calls these types of tools “visual COBOL” and I think that’s spot on.

udidahan Says:
July 14th, 2011 at 12:46 am
Tommaso,

You might want to look at my blog post on integration where I describe how to do it without the use of an orchestration engine:

http://www.udidahan.com/2011/04/08/integration-how-and-where/

udidahan Says:
July 14th, 2011 at 12:48 am
gigiabbrescia,

Even if you take the visual parts out of the equation, centralized architectures tend to lead to more monolithic code bases – less maintainable and worse performing than distributed architectures.

udidahan Says:
July 14th, 2011 at 12:50 am
Jimit,

Distributed pub/sub architectures don’t preclude workflows, but don’t work with *centralized* workflows. Within each publisher or subscriber, there could be some workflow. It’s been my experience that visual tools don’t help under most circumstances.

udidahan Says:
July 14th, 2011 at 12:54 am
David,

On the practical side of things, I’d prefer to use an architecture where the tooling enables me to get things done productively, resulting in high performance and maintainable code.

If a centralized workflow tool was created that handled all the concerns I raised above, I would evaluate it on its merits.

Regardless, centralized architectures have proven themselves problematic repeatedly leading us, as an industry, to things like the Single Responsibility Principle – dividing things up so that they don’t all sit in one place.

Anyway, if you hear of such a tool – ping me. I’d love to take a look.

Charlie Barker Says:
July 14th, 2011 at 1:46 am
Managing complexity is the fundamental challenge. If we let any part of our system get into a state of high complexity without putting the neccessary controls in place we end up in a world of pain.
One of my battle scars was a system that was highly configurable, the problem was this meant administrators were overwhelmed and the task of configuration of all but the simplest of settings was handled by developers. Configuration is harder to test than code so in the end it was a slower and less reliable way of introducing change to the system.

Nuno Lopes Says:
July 14th, 2011 at 4:00 am
Humm. I’m not sure if I fully agree with you assessment Udi.

I know what you mean as I’ve seen horrible things happening with visual tech. These usually started using false premisses:

1) Workflow IDE are for business people (analysis and so on)
2) Workflow IDE require minimal to none programming experience

The fact is:

1) Workflow Visual Tools are for programmers. They are mainly “code” generators (accelerators)
2) Workflow Visual Tools require some experience in logic and programing.
3) They allow analysts to communicate better with programmers and vice versa.

The problem is that vendors push these products to top level analysts, when indeed they are all about development productivity. This leads to all sorts of problems.

J-F Declercq Says:
July 16th, 2011 at 1:37 am
The main underestimated impact of workflows is exception management. If you ask yourself ‘what if somthing goes wrong’ on each worflow step, you’ll maybe understand what I mean. Compensation, roll-back…

You need to have th possibility to leave the main flow and propose user functions which allow either to put the instance back in the flow or to execute the tasks in a different way…

Therefore you need to consider the right combination of portal, bpm, bam, rules engines, information management and services in function of each workflow.

tom Says:
July 16th, 2011 at 3:02 am
The only motivation I can think of for visualizing code like these tools do, is to have people maintain the logic who do not have the ability to write code.

I dont see why these tools should be appealing to programmers. For me they have no appeal at all.

udidahan Says:
July 16th, 2011 at 11:10 am
Charlie,

I’m sure, at the time, the idea was that people other than developers would be the one doing the configuration. The road to hell is paged with good intentions

udidahan Says:
July 16th, 2011 at 11:13 am
Nuno,

I’m not sure that I’d agree with the assertion that these visual *programming* tools help business analysts and programmers communicate better. I’d say that verbal communication focused on “what should happen when we don’t get a response from X in T?” will drive out what the real “process” should be.

udidahan Says:
July 16th, 2011 at 11:16 am
J-F,

It’s been my experience that the majority of logic ends up revolving around “exception management”. While it is possible for an experienced developer not to be “duped” by the things that these visual tools make easy, the question is whether those tools are appropriate once you realize the non-sequential nature of the world.

udidahan Says:
July 16th, 2011 at 11:18 am
Tom,

Well, these tools *do* demo well and help convince a CIO to choose one platform over another.

Matt Says:
July 18th, 2011 at 8:08 pm
I agree the tooling for workflow engines is terrible, but it does not mean the idea is bad.

The fundamental problem is that modeling the workflow surfaces complexity otherwise hidden until it resurfaces later as bugs or missing features. This makes the problem appear harder than it actually is.

Modern tooling does not help the matter. Many tools are graphical and sold on the idea that laymen can use them. Trying to represent complex process (a N dimensional graph) graphically (2D) is a very hard problem. Techniques such as AOT or DSM can be used to subdivide the graph into a lower dimensionalities but these are advanced and rare.

In my experience the most successful workflow engines use Monadic Combinator Libraries. If the domain is narrow enough you can bolt on a DSL / DSM tool on top. And these can finally be made simple enough for a laymen to use them.

So all you really need is really good developers

- Matt

verticonaut Says:
July 18th, 2011 at 11:31 pm
The diagram looks scary – right. What’s missing is the ‘handwritten’ code that would replace it. I’m pretty sure that would look scarry as well! Especially when written by different developers.
Where I agree is that a workflow tool doesn’t help when a workflow is complex. It will be hard to develop and maintain in either way. And for the rest of the ‘not so conplex’ ones a workflow tool might not be needed at all.
In large companies, ‘centralized’ and ‘tool or frameworks’ are often considered because they strrugle with inventory and documentation of their workflows. So the decision of using them not really driven by the value of this tools in development – which is limited as you say.

Jerry Bullard Says:
July 19th, 2011 at 4:28 am
Apples, meet oranges. You seem to be comparing workflow for everything and code for everything, when in reality each can have a role in your architecture. Nothing prevents us from implementing small, autonomous workflows as publishers and subscribers in a SOA/EDA. Your primary argument then boils down to “monolithic things are bad.” Agreed.

Also, the notion that “real coders use text” is just silly bigotry. You are ignoring the capability of workflow to be an extensible domain specific language (speaking primarily from the .NET WF4 perspective) with baked in best practices. Ultimately code is code and requires a developer mindset, but we all seek higher levels of abstraction for productivity (or to provide that abstraction to some customer without quite the same level of programming and debugging abilities as we have). You can use workflow to let your customers extend your system while providing the rails and bumpers to keep them from crashing into the walls. Likewise you can use workflow to extend someone else’s system. Beats learning some kind of macro or specialized scripting language.

Saying it is difficult to write a test for workflows ignores the ability to test each activity in the workflow independently. This is a little like saying it is hard to test a programming language or everything anyone could do with an API. Likewise it is difficult to test a SOA/EDA as a whole. The universe can take on an infinite number of states, so we do our best to test the transitions.

Finally, not all workflows are simple flow charts. Check out state machine workflows in .NET WF, and you should see a way of managing complexity and visualization of a process out of reach for many, while fitting perfectly in the SOA/EDA ecosystem.

Giscard Says:
July 19th, 2011 at 7:30 am
Was going to write some of what Jerry Bullard said.

And I kind of disagree with the premise of your post.

“The main touted benefit of these workflow-centric architectures is that we don’t have to change the code of the system in order to change its behavior resulting in ultimate flexibility”

Pretty sure this is a well known fallacy. The value add as a whole is what sells these products. The designer is a pretty small piece.

Sadly your post is still a fair caution to many.

udidahan Says:
July 19th, 2011 at 11:58 pm
Matt,

I’m willing to agree that what you really need is good developers – they then will be able to decide which tools/techniques should be used where.

udidahan Says:
July 20th, 2011 at 12:00 am
verticonaut,

I think you hit the nail right on the head – for big complex things, these workflow tools add to the complexity; for small things, those same tools are unnecessary.

udidahan Says:
July 20th, 2011 at 12:09 am
Jerry,

Seeing as you agree that centralized/monolithic things are bad, and that many of these tools are evaluated based on that exact usage scenario, I think we’re not that much in disagreement.

After dividing up a system according to SOA/EDA, one tends to find that most traditional workflows no longer exist, rather they are the result of cascading events from various publishers and subscribers.

What is left are sagas – a kind of workflow for which some tooling could be helpful. While one area where a tool might be useful is for integration sagas, there are other solutions that tend to be better (like I wrote about here: http://www.udidahan.com/2011/04/08/integration-how-and-where/ ) Another area where sagas are particularly valuable has to do with implementing business policies (a special kind of domain model) and these tend not to be either sequential or necessarily suited to a state machine. A rules engine would be a stronger candidate for a tool, but often the tooling gets in the way of the developer understanding the problem.

udidahan Says:
July 20th, 2011 at 12:10 am
Giscard,

Even things which are well known to some need to be repeated for others.

Larry Lynch Says:
July 20th, 2011 at 3:57 am
I hate PEGA
we are using PEGA wasting millions of dollars where we could have written code for a week to achieve the same thing with testability, mockability, readability and predictability.

steve Says:
July 20th, 2011 at 4:10 pm
I agree with the notion of decentralized Sagas especially for business policies – but your link to integration-how-and-where is broken. Have you seen any tools that fit this mold?

I’ve put FluentWorkflow into production apps and it works extremely well – no underlying tooling/runtime required to make it work. Curious if you have seen others.

Juergen Aschenbrenner Says:
July 21st, 2011 at 2:22 am
Good point Jerry Bullard made,… its like the hammer nail problem found in many other problem domains: One has learned to use a hammer and soon every problem looks like a nail.

I like BPEL and other visual programming tools, because when properly used help gathering requirements from business experts. Same as with mockups when designing the UI.

The problem with long running batch processes prior to visual tooling, was that there weren’t any artefact’s one could use as a basis to start discussing problems at hand.

What part of problem domain should be represented visually and what with traditional programming languages should be at the discretion of the developer. Tools like Websphere ID just allow to do that.

Yoni Goldberg Says:
July 21st, 2011 at 2:45 am
The business motif: While working with my customers (usually enterprises), I noted that nothing is more important than business agility, my customer understand their business better every day and tends to asks for changes on a regular basis. Sometimes, when code is involved, small changes might be implemented in two weeks or maybe a month, while waiting so long for small changes – the customer loses his patience and belief in the program. Contrarily, when changes are committed really fast, the customers tends to keep asking for changes frequently which makes them see the software as a reflection of their business.

I believe, especially in enterprises, that nothing drives agility more efficiently than the existing of implementer/power user who works closely to the business user and performs all the changes for him. Workflows tools allow implementer and Power users to change business process rapidly. Code changes involves longer IT procedure for approval: every change requires to make a contact with the IT department (or software vendors), wait for availability, ask the admin for permission to change code on the servers, install on testing servers and more – I see how those procedures exhaust my customers. All the goodies of Workflows agility can live together with robust architecture which handles complexity very well.

The architecture: Workflows are never meant to handle software complexity. The complex parts, as always, should be encapsulated in software units (for example, DDD entities which are consumed using services), WF activities consume the complex business entities, and then the workflow itself merely an orchestration of software activities which hide the complexity very well.

After delivering WF activities for the WF system, the implanter can craft multiple different business processes using those activities, He can make changes really fast, change the sequence, change parameters, change HIGH LEVEL if-else conditions and make many other changes. I see at work, how developers code once some software component which then is being used in many different ways by implementers using WF tools. Of course, we cannot change everything using WF tools, sometimes we have to change inside the software entities but it gives a layer of agility.
So, IMHO, in large systems, workflows are not supposed to be the software itself neither the business logic layer but rather a part of the application layer or facade which consumes software components in techniques that drives agility.

udidahan Says:
July 21st, 2011 at 8:39 am
Larry,

But you’ve got to hand it to PEGA – they’ve made a killing selling these kinds of all-singing-all-dancing tools

udidahan Says:
July 21st, 2011 at 8:43 am
Steve,

I’ve fixed the link. I’ve never used FluentWorkflow but from a quick look at it, don’t like that it makes states so prominent. NServiceBus sagas allow you to use timeouts to “throw state into the future” without having to model it explicitly. These kinds of lightweight solutions often lead to simpler code.

udidahan Says:
July 21st, 2011 at 9:54 am
Juergen,

My concern about most of these BPM tools is that they’re usually not very good at representing time, or that they don’t guide the discussion in that direction. In that sense, leaning on a tool to facilitate discussion between IT and business may be counterproductive.

udidahan Says:
July 21st, 2011 at 9:58 am
Yoni,

Distributed architectures can greatly facilitate the exact same agility you mentioned by bringing down complexity, enabling to upgrade the system piece by piece, while still maintaining all the robustness and testability of business processes (which ARE logic, no matter what anybody tells you).

Wes Says:
July 22nd, 2011 at 12:21 am
Amazeballs, best synopsis of this topic yet, thanks!

nightwatch Says:
August 22nd, 2011 at 12:27 am
It’s great you started a discussion about workflows here, but I’m afraid you jump to conclusions too quick. Probably because you already have the answer for questions not asked yet: NServicebus
First of all, a BPM engine can and should make use of SOA and services provided by applications and it can make use of a decentralized message bus. Business processes usually go ‘through’ several applications and the software must reflect that by providing the means of coordinated data exchange. You can split the process logic between these applications but without any distinct coordinator you end up with fragmented logic that is difficult to maintain in almost every aspect (versioning, development, production rollout, testing, prototyping, documentation, vendor coordination, …).
With the right BPM tool you can make a great use of your existing applications and services they provide, my experience is that BPM adds a great value if you can make it work with existing applications/services. The most important benefits, imho, are:
- versioning: co-existence of multiple process versions is possible
- easy modfication of process logic, very often without affecting applications involved
- documentation: process definitions are self-documenting (mostly)
- common standards for logic description
- faster development, testing and time to production
- you can hand more responsibility for the business logic to your client (so they can understand and even modify it without paying the vendor)
- and many more. Please note I’m not a salesman but rather a software architect with some experience in analyzing and implementing various business processes in telecoms and other big organizations. I’m also an author of a simple BPM engine that proved a very useful tool exactly because of features mentioned above. I agree that real-life process logic is very complicated and building a process description in some modelling language is just programming, not drawing, but please dont try to convince us that you can avoid the complexity by scattering the logic between multiple applications.
And your last argument about testing,debugging and versioning: why do you think process definitions can’t be versioned or diffed? They are normal text files with structured code/data inside, not GIFs, PDFs or handmade drawings. Testing and debugging is also possible, and can be quite easy if the BPM engine allows you to look inside process state.
Sorry, I must be going now, hope this subject will be discussed more here.

udidahan Says:
August 31st, 2011 at 11:26 pm
Nightwatch,

With regards to your statement:

“You can split the process logic between these applications but without any distinct coordinator you end up with fragmented logic that is difficult to maintain in almost every aspect…”

I don’t recommend dividing up this logic between applications but rather between services. Also, in my model of SOA, services aren’t provided by applications. Instead, one can think of an application is a hosting process for pieces of multiple services.

On your statement:

“my experience is that BPM adds a great value if you can make it work with existing applications”

The philosophy of integrating existing applications (without changing them) has been tried before – it’s called Enterprise Application Integration (EAI). This led to spaghetti integration logic implemented in EAI brokers. Now this exact same philosophy is being touted under banners of ESB and BPM. Lipstick on a pig, as far as I’m concerned.

I would be interested, though, in seeing some examples of a diff/merge process when multiple developers are working on the same set of processes. If you also have an example of a test of a process that calls in/out to various apps, I’d like to see that as well.

Thanks for your comments!

nightwatch Says:
September 7th, 2011 at 4:03 pm
Udi,
I’m trying to continue this discussion but the subject is so broad that I’d have to spend all day writing an answer. So instead I’ll just concentrate on your statements:

>> I don’t recommend dividing up this logic between applications but
>> rather between services.
I fail to see a difference here. In my world applications provide various services from a business perspective but these services are a product of the underlying logic, not the other way round. Maybe you are talking about services built ‘in between’ that provide the logical and technological glue between applications, or the missing parts of process between them?

>> The philosophy of integrating existing applications (without
>> changing them) has been tried before
No, I didn’t say ‘without changing’. Just without adding a logic responsible for something that should be outside the application, like binding to other systems or data transformation required by integration.

>> I would be interested, though, in seeing some examples of a
>> diff/merge process when multiple developers are working on the same
>> set of processes
I haven’t seen multiple developers work simultaneously on multiple processes. One process is sometimes too much for an IT project. Regarding the technical side, if the process definition language is concise and no automatic code generation tools are used, I see no problem – merge just like ordinary code.
And spaghetti code imho is the result of not separating the logic of different processes or different versions of the same process – this way the changes accumulate until they obscure all the logic.

>> If you also have an example of a test of a process that calls
>> in/out to various apps, I’d like to see that as well.

My approach to building and testing processes is different. Generally, the IT role in process implementation is to automate otherwise manual tasks. And the automation is a gradual process – you identify most important tasks that need to be automated first and leave the less important to be done later. When building your process definition in a BPM tool you start with all manual tasks, then you make sure the process model is accurate, and then you start replacing them with their automated versions, one by one. The most important thing is that the overall logic and data structures shouldn’t change, you just replace a manual task with its automated version without changing its interface to the rest of the process (by interface I mean input and output data). This way the tasks are completely interchangeable, provided that the interface is kept unchanged. And for testing you can ’simulate’ any external system with a manual task or a ‘mockup’ code that responds with the same data structure.
That’s why I said it’s very important that BPM tool allows for easy changes in the process logic

David F Says:
June 23rd, 2012 at 1:02 am
Hi Udi,

I think Yoni makes a reasonable point. I agree distributed architecture encourages loosely coupling between components, but my experience showed that distributed architecture can also downgrade the system performance greatly unless using much more expensive hardware solution.

So, my point is that a proper combination of centralized and distributed architecture should be adopted based on a careful analysis.

Please correct me if I am wrong.

Thanks!

Dave

udidahan Says:
June 24th, 2012 at 4:31 am
Dave,

Well, there’s the first rule of distribution: Don’t.

So, in some sense, you are correct. That being said, there are ways to analyze business processes so that they can be broken down into more parallel pieces that are triggered off of various events.

In this model, distribution does provide performance gains as we can have more machines doing more work in parallel, but it would also scale well with more CPU cores on a single box as well.

Hope that makes sense.

Sandeep Says:
December 20th, 2012 at 4:57 am
Though what you say makes some point, I dont see that you know the centralized tools you mentioned use the pub/sub architecture at the core and it depends on us how effectively we utilize it. By going with your approach of writing your own code to create a pub/sub architecure is like reinventing the wheel which someone has already invented. I am not totally convinced with the coding aspect for such scenarios. Instead of writing own pub/sub architecture I think, thinking process should go in how effectively we can use the on the shelf tools available.

udidahan Says:
December 25th, 2012 at 10:48 pm
Sandeep,

I’m not recommending rolling your own pub/sub tooling. What I am saying is that when choosing a tool, look for one that doesn’t lead you to create centralized workflows. NServiceBus is an example of that kind of tooling.

Robert Says:
January 17th, 2013 at 12:05 pm
First graph looks nice. Much easier to analyse than most ndepend graphs of object oriented projects I’ve worked on.

Gilligan Says:
May 7th, 2013 at 9:52 pm
I know this is late in coming but, I have been workign at my current company for 4.5 months and when I first came in and saw what they were trying to build I teamed up with the other programmer on the project ot show the boss what a monumental bad idea it was. And the idea was exactly this. It was a workflow engine that they wanted to have a “user-interface”. The boss agreed with us and we quickly scrapped the idea. And we have every day been appreciating how much work that saved us and how cleaner our workflow stuff now looks.