camunda bei der Neujahrsfeier 2012

2012 ist schon nicht mehr ganz so jung – aber für alle, die sich für camunda generell interessieren, kommen jetzt ein paar Status-Infos zum Stand unserer Unternehmensentwicklung:

In 2011 haben wir erstmalig mehr als eine Million EUR Umsatz erzielt (sogar deutlich mehr). Das macht uns natürlich stolz, zumal wir nicht einfach “irgendeine Unternehmensberatung” oder “irgendein IT-Dienstleister” sind, und ständig “irgendwelche Consultants” in Projekte stecken, nur um kräftig Tage zu fakturieren. Stattdessen fokussieren wir uns nach wie vor zu 100% auf BPM, und treten viel mehr als “Enabler”, also “Befähiger” unserer Kunden auf, denn wir geben konzentriertes Know-how zu einem abgrenzbaren Themenkomplex weiter. Dementsprechend kurz sind in der Regel auch unsere Einsätze, und dementsprechend breit ist unsere Kundenbasis: Insgesamt haben 254 unterschiedliche Unternehmen und Behörden in 2011 unsere Leistungen in Anspruch genommen, wovon 48 Neukunden waren.

Aber natürlich gab es auch Projekteinsätze, die wir vor allem deshalb durchführten, weil wir bei spannenden und innovativen BPM-Projekten gerne am Ball bleiben und daraus wiederum Know-how für unser “Premium Consulting” ziehen. Zugenommen haben in 2011 auch die “begleitenden Coachings” auf Basis eines punktuell abrufbaren Kontingentes, wie es zum Beispiel Energie Südbayern für die Einführung von BPMN in Anspruch genommen hat (siehe Praxisbericht).

Neben den Themen BPM im Allgemeinen und BPMN im Besonderen hat in 2011 auch ganz klar das Thema Activiti zugenommen: Deutlich mehr als die Hälfte unserer Einsätze im Premium Consulting und der Projektarbeit bezogen sich auf diese ja noch vergleichsweise junge Open Source BPM-Lösung. Wir merken also sehr deutlich, dass Activiti in der Praxis angekommen ist und im Grunde einen ähnlichen “Siegeszug” erlebt wie die BPMN – wenn auch im kleineren Maßstab. Besonders spannend ist für unsere Consultants auch der Einsatz im Ausland, der dank Activiti in 2011 bereits spürbar zugenommen hat (innerhalb Europas, aber auch weltweit).

Da der Einsatz von Open Source beim Kunden natürlich einige Fragen aufwirft (Support, Stabilisierung, aber auch fehlende Features) ist es kein Wunder, dass wir diese Fragen zunächst punktuell, inzwischen aber auch in einer vollständigen Lösung beantwortet haben: camunda fox wurde geboren. Wer sich für die Aktivitäten von camunda in 2012 interessiert, findet hier auch bereits die primäre Antwort .

Mit dem Auftragseingang ist natürlich auch unsere Mannschaft gewachsen, die inzwischen 16 Köpfe zählt und Anfang Januar ihre traditionelle Neujahrsfeier zelebrierte (siehe Foto, wenngleich vier Kollegen leider nicht dabei sein konnten). Wir sind also – ganz bewusst – nach wie vor ein überschaubares Team, und ich denke in mancherlei Hinsicht auch ein einzigartiges: Hervorragend in unserer Kernkompetenz BPM, sind wir gleichzeitig unseren Werten verbunden. Viele der scheinbaren Sachzwänge der traditionellen Geschäftswelt werden von uns schlicht und egreifend ignoriert, und wir sind nicht trotzdem, sondern gerade deshalb erfolgreich.

Wer also BPM liebt und gleichzeitig den Mut hat, neue Wege zu gehen, damit eine Organisation besser funktioniert, den laden wir herzlich ein, uns auch in 2012 zu begleiten: Als Kunde, Partner oder Mitarbeiter. Gemeinsam werden wir eine Menge bewegen

Both the Activiti community and our consulting customers are repeatedly asking me about how to run Activiti on IBM WebSphere Application Server. In this blogpost I want to summarize the problems and challenges you face when you want to do that and along the lines give a sneak preview of the upcoming WebSphere support in camunda fox (our enterprise BPM platform based on Activiti).

Let’s start with the challenges.

Transactions

A first challenge is transaction management. When developing a process application with Activiti, you often want the process engine to participate in container transactions. This allows you to implement units of work in which both the process engine and other transactional resources like a database, a message queue or EJB’s participate. The Java EE Solution to this problem is JTA, a specification which defines multi-resource transactions. Activiti integrates with JTA out of the box, if you configure the JtaTransactionInterceptor. This makes sure, that whenever you run a command (do something in Activiti, like completing a user task), we either start a new transaction or participate in the currently active transaction. The desired behavior is that if you invoke Activiti form an EJB Session Bean, it will take part in the transaction managed by the Session Bean / EJB Container. The same is also true in the opposite direction: lets say you call an EJB from an asynchronous service task, in that case Activiti will open a transaction when executing the service task and the EJB has the possibility to participate in that transaction. Why is this so helpful? As I said before it gives us the possibility to implement units of work in which both Activiti and the EJB Container take part and by the means of transactions ensure consistency: if Activiti fails, the EJB fails and vice-versa (remember ACID transactions?).

Now, why is this challenging on WebSphere? WebSphere Application Server does not provide a standard JTA javax.transaction.TransactionManager (like for instance JBoss AS or Glassfish), but instead exposes a proprietary interface named com.ibm.wsspi.uow.UOWManager. IBM is allowed to do so, as the Java EE / EJB specification does not require vendors to provide a JTA TransactionManager implementation. All they are required doing is providing a javax.transaction.UserTransaction (which Websphere does). The UserTransaction interface would allow Activiti to open new transactions and participate in existing transactions; however that is not enough for Activiti to function correctly. For some commands (for example when decrementing retries after a failed job) we need REQUIRES_NEW semantics for executing the command. This means we need to suspend the current transaction and start a new transaction. Now the javax.transaction.UserTransaction interface does not provide such functionality, which means that you need to use proprietary IBM API for achieving this. This is one reason among (others) why running Activiti on WebSphere application server requires you to solve the transaction problem. One possibility would be to use Spring’s org.springframework.transaction.jta.JtaTransactionManager abstraction and configure the SpringTransactionInterceptor in Activiti.

Performing Background Work

A process engine like Activiti uses the calling thread for performing most of its work. This is a natural consequence from the requirement to share transactions between Activiti and the application code, as transactions are thread-bound. This means that, when you call Activiti, the call will block until Activiti has performed its work. But there are also certain situations where Activiti will call you. For example if you have a timer in the process, Activiti will wait for the timer to reach its due date and then continue with executing the process. Hence Activiti performs some kind of background work, where it checks if it has work to do (the concept is called “Jobs”) and if so, it will perform this work. To make this happen Activiti needs to manage its own threads.

Now, in a managed environment like an applications server, you do not want to manage the threads yourself, instead you want the container to manage them (more exactly: it is even forbidden by the Java EE Specification). This is for various reasons like for instance to allow the container to perform additional “magic”, like monitoring the threads, detecting deadlocks, ensure that work submitted by an application is canceled when the application is undeployed, that threads requested by an application are stopped if the application is undeployed and so on. Most of this is to ensure that “everything is under control” and to ensure the overall stability of your application and server environment. Another important feature is ensuring that work submitted by an application is executed within the correct security / naming / transaction context.

When running Activiti on WebSphere, the main challenge is having the JobExecutor (this is the component which performs the background work) delegate to supported, container-managed threads. On WebSphere this means delegating to an asynchronous beans WorkManager. There are multiple interfaces to the WorkManager; one of them is commonj.work.WorkManager, another one com.ibm.websphere.asynchbeans.WorkManager. Using the WorkManager for Activiti background work was a problem until recently, as the activiti JobExecutor always started its own Threads and it was not easy to configure it. Recently we have done a number of refactorings in which we separated the JobExecutor logic from the thread management, such that it is now easier to implement a JobExecutor delegating to a Websphere WorkManager (watch out for Activiti 5.9).

Ok, so if we know all this, why is there no Activiti distribution for WebSphere you can just download?

Why is WebSphere Support not pre-built in Activiti?

The reason is that it is not only a technical problem. Activiti aims at being a highly configurable and embeddable engine. This means that conceptually it possible to integrate Activiti with whatever Java-based environment you come up with. This is a good thing! However, that also poses some challenges. One challenge is knowing where to stop. Take for instance WebShere Support: let’s say we add WebSphere support classes to Activiti. Now the main problem is to test and to maintain it. This includes automated testing and quality assurance, preferably on different versions of the application server, different Java Runtimes (e.g. the IBM JDK) and so on. That kind of effort is impossible for us to do in the context of an open source project, especially as WebSphere is not the only application server out there. This is why we focus on delivering a high quality embeddable process engine in the Activiti project, get that right and do not assume responsibility for integration with various platforms and vendors. Because, as I said before: even if we made that promise, it is hard for an open source project to deliver on it. And don’t forget: Somebody has to pay my salary

One way out for you by the way could be the Spring Framework which already contains abstractions for many vendor-specific technologies like transactions and threading. So by providing first-class integration with the Spring Framework, we give users of Activiti the possibility to integrate with various vendors more easily, but it still means you have to do some integration work yourself and especially take the responsibility for the integration and maintenance of it yourself.

Support in camunda fox

In the camunda fox enterprise bpm platform we provide pre-built integration with IBM WebSphere Application Server 8. We do perform automated testing, quality assurance and maintenance and we offer professional support. This is one of some differences between Activiti open source and camunda fox enterprise.

While this is no “official” announcement about the camunda fox platform for websphere, I cannot stop myself from giving you a sneak preview about what you can expect in terms of functionally. You must have already gathered some of it form this blog (I know, it’s long again…): we support container managed transactions and container managed threads for background work. In addition, we support a central process engine service which can be shared by all applications installed on the application server. That also means that it allows you to align the container deployment lifecycle with process deployment lifecycle, i.e. when you deploy a WAR file (application) which contains processes, the processes are automatically deployed to the central process engine service. This is includes some class-loading magic as well. But more on this at a later point in time. So for the moment, enjoy the following screen-shots:

Process Engine Configuration using the WAS web console

Configuration of the WAS WorkManager used by the JobExecutor

The demo application I deploy there is the one I wrote for the last WJAX, see my blog post with a screencast including the whole presentation. If you are quick I have not yet deleted the cloud instance, so check it our yourself: http://orderdemocloud-camunda8.rhcloud.com/order-confirmation-rules/.

Cycle Openshift Plugin

• Private Clouds: Think of the same mechanism inside your company. No longer long processes required to get a test or integration system up and running (how long does this take at your company in average? I saw even months at some customers, including hardware procurement).
• Quick-Start: For some companies the data security aspect of having the application running in the cloud might not be a major drawback, especially compared to the advantage of being fast and easy. Not to mention, that the starter edition of JBoss OpenShift is for free, at least at the moment.

In order to get started:

• Create a SSHkey pair (yeah, this is unfortunately needed to get it all running), I used PuttyGen (remember you have to export a RSA key and this must be copied to ~/.ssh/id_rsa).
• Register for JBoss OpenShift
• Get a GIT environment up and running on your local machine (e.g. Putty & TortoiseGit on Windows)
• Download the cycle preview
• Checkout and build the Openshift-Plug-In for cycle by yourself and add it (and all necessary third party libraries as e.g. JGit) to the classpath of cycle (basically the WEB-INF/lib folder)

Please remember: This is a proof of concept, the plug-in was hacked on one Saturday and is not production ready! We just wanted to demonstrate how easy it is to write plug-ins and how cool use cases you can implement I have some things in it only working under windows which would need to be fixed (e.g. calling “mvn.bat”). So watch out for the the TODO’s in the code if you want to play with it or maybe even use it in real projects…

Finally we camunda business people have a ‚success story‘, too. In this case it is about a BPMN modelling project. In the context of the Modernised Customs Code, we modelled the export processes together with the German Federal Ministry of Finance and MATERNA GmbH. Camunda was responsible for all questions around BPMN.

Most models were independently modelled by the employees of both enterprises after the first workshops. It was remarkable that even the most complex models were discussed in joint workshops with all participants. Lawyers, representatives of the business side as well as IT specialists were able to clarify their questions together – a very useful approach but unfortunately far too seldom used in practice.

Enjoy reading: Download

Activiti & Drools in Action

order confirmation process and surrounding process solution artifacts

Drools Regeln werden aus Entitäten der Regeltabelle erzeugt

You can download the camunda fox server and the example application. Just adjust the path to your JBoss in the build.properties file, then you can use the ant file for deployment. You can access the application by this URL afterwards: http://localhost:8080/order-confirmation-rules/. Have fun!

Nun endlich, wie versprochen und von vielen schon sehnsüchtig erwartet, die Folien zu den Vorträgen! Die Folien zu den einzelnen Vorträge, soweit sie zur Veröffentlichung freigegeben wurden, finden Sie hier:

‘Begrüßung’ Jakob Freund, camunda

Keynote: ‘BPMN & Co. Status Quo and Perspectives of PMGs Activities around BPMN’. Andrew Watson, Vice President & Technical Director, OMG

‘BPMN im Projekt ATLAS Ausfuhr des Bundesfinanzministerums’ Eileen Hildebrand, MATERNA

‘BPMN und UML im Requirements Engineering’ Martin Schimak, Berater

‘BPM(N)-Softwaremarkt: Überblick + Tipps zur Tool-Auswahl’ Jakob Freund, camunda

‘Die OCEB-Zertifizierung für BPM-Experten’ Matthias Schrepfer, camunda

Dawn at 300 km/h

On the other hand it is a good thing to distinguish between structured processes and those that are more or less unstructured and therefore not really suited for the taylor-oriented management approaches, let them be ways of organization or certain technologies like process engines. I just do not like the connotation that “the world” becomes more and more unstructured, or that taylorism is somehow “evil”, because all those statements sound like ideology to me, and ideoloy has never been a good advisor.

I would even claim that

a) taylorism as a good thing and
b) The limit of taylorism in the world of services is by far not exceeded and therefore
c) taylor-oriented technologies (like process engines) have a damn bright future

Huh, how could he claim THAT!? Tailyorism, isn’t that the thing with the machines and the inflexible production lines and all those zombie-workers that are completely unhappy in their daily work? Well, more or less, yes. But there is one completey killing argument: It is also the one that scales.

Let me explain it somehow different. There is a really nice quote I recently read by Alfred North Whitehead: The progress of civilization can be measured by the quantity of actions we can execute without thinking. And I also have a really nice example: Traffic lights. When I approach a crossing driving my car, and the lights are green, I just go on driving, with about 50km/h in German cities. This is rather fast, but I am completely sure that there will be no one coming from the left or right crashing into me, because there is a defined system I can rely on.

Actually, when I am driving, I am a zombie worker most of the time. Sometimes, of course, there are “unpredictable” events, like a child running over the street, or an alien spaceship landing in the middle of the highway. Then I become a knowledge worker, handling that case with my horribly flexible brain.

But what if I would be a knowledge worker all the time, if there would not be any traffic lights (or traffic rules in general)? I would approach every crossing, looking right and left and right again, and then carefully cross it. I would do that *all the time*! What does that mean for my grade of efficiency, and for that of all the other “knowledge workers” in German streets? If German tourists drive in Italy or Spain, they are often completely scared of all the people ignoring the traffic rules, of all the *unpredictableness* on those streets. Actually I think, if it comes to driving, Germans are more like zombie-workers, and people in other cultures are often more like knowledge-workers. But hey, you can say anything you want about Germans, but for sure we are rather efficient.

So the bottom line is: Making the world more predictable (yes, it can be done), and then applying axiomatic systems to it, is nothing invented by taylor and somehow an “accident” of the 20h century, but it is a central component of human evolution. It has always been there, and it will always be there, as long as people are interested in less work and more free time.

OK, so how do those thoughts apply to BPM in general and process engines in particular? They are about axiomatic systems for business processes, about standardization and less flexibility, but also about less work and more free time, or in other words, less costs and more profit: scalability. If I am running a business, this is my biggest interest. If I need knowledge workers for those parts of the process that are unpredictable but necessary (!), so be it. But my ambition will always be to reduce those parts, because knowledge workers are expensive, and they cannot be copied, and therefore they are not scalable, and therefore I consider them a risk and the need for them a problem I have to solve sooner or later.

OK this sounds a bit harsh, and of course it is not true for every kind of business. But I want to explain why there is still so much to do for BPM and process engines in the world of services, like insurances, banking, telco etc. One example: In Germany we have a lot of rather old, traditional insurances, and we have the new ones that operate their marketing and sales activities mostly online. Those new players could realize their processes on green fields, they used a lot of BPM methods and technologies without the need to regard existing culture, policities, habits and people who must be kept busy because you cannot fire them (and who therefore are considered “knowledge workers”). Those new players are small and extremely efficient, and in those market segments where you can already apply that kind of taylor-oriented approaches (like car insurances), they are a real threat to the traditional insurances.

So I think while the discussion about unstructured processes and how we can handle them is a good one, it is a bit exaggerated. In the end, optimizing business is about scalability, and scalability means production lines, and that means taylorism. Case Management won’t change that.

But I also think that most of the existing BPMS-Technologies lack certain abilities to realize good process applications, which means besides other issues, those that can cover both structured and unstructured parts. At camunda, we aim at those issues with a new kind of BPM-Platform called camunda fox, based on the open source project Activiti. There have already been a couple of blog posts about it, but those adressed mostly IT-people like architectes and developers. In my next post, I will explain what makes fox different in a way that it is also interesting for those of us that are not particulary interested in Class Loading, Persistence Layers and the like

But step by step. JBoss 7 is now certified for the Java EE 6 standard and is now the main server (which will be productized as well) of JBoss (so keep your fingers away from JBoss 6!). JBoss 7 was completely re-factored internally and is pretty much modularized. This allows a really impressive startup time of only a few seconds of a naked server and even with a mid size application 10 to 20 seconds. It only starts the modules it needs and can even load modules on-the-fly as soon as an application is deployed which needs them, which is pretty much comparable to what OSGI does.

With that in mind it now is really a good questions when and why to use Tomcat instead of JBoss 7, since a lot of the main arguments are gone (Startup time, complexity, …). But this is a completly other story I don’t want to discuss here today. And Tomcat has announced TomEE, a Java EE 6 compatible server as well

But this leads to another interesting aspect: JBoss 7 is currently certified only for the WebProfile of Java EE 6. Other important features like Remoting (sometimes necessary for EJB applications ) are not yet implemented. But actually JBoss is moving pretty fast at the moment, the JBoss 7 is strategically very important for them and it gets a lot of recognition out there, so I expect them to deliver a complete EE stack soon.

In short: JBoss 7 is really amazing and I am impressed how stable the current version already is. And we didn’t had problems in production yet The Java EE 6 Stack together with the camunda fox and Activiti CDI Integration is cool and it makes a lot of fun and not much time to implement “process applications” For more information check the links below:

• camunda fox development stack
• “Less-Code BPM” with camunda fox server, Activiti and JBoss AS 7
• Build your own activiti task explorer with CDI and JSF 2

So ne excuses any more, start with JBoss 7 now! Download the vanilla server (at JBoss) or the camunda fox server with Activiti already integrated as a preview (at camunda). Let us know what you think!

So, enough blogging, no I get back to write a quick prototype with JBoss OpenShift to deploy a process application with one click in camunda fox cycle in the cloud (Amazon EC2), e.g. for testing it. Let’s see