How Charming is Laziness?

Long time ago, while I was still a student at the PolyTechnic in Enschede, The Netherlands, I would justify my choice to study computer science, by proclaiming that laziness leads to efficiency. Which of course makes no sense I now know, because it leads to effectiveness. Penny-pinching leads to efficiency. You can read all about it in this post Perish or Survive, or being Efficient vs being Effective. But there's this little concept that results in more efficiency and laziness as well. It's therefore a charm. It is called automation and closely related to that third time.

Third Time Is Automated Principle

One of the key principles at Amazon AWS is that everything must be automated. It's not just that everything should be automatable, but it should be automated.

Whether or not it's an urban legend, but word has it that when you create something that requires manual action, you're out looking for a new job. For a company like Amazon AWS it is clear why automation is such a huge thing. And it is clear as well why they have such focus on API's. API's are how automation across the board is facilitated. But most companies are not Amazon or any of the other cloud providers. Most of you my dear readers are more likely to run your systems on Amazon AWS, Google Cloud or Microsoft Azure, then run those applications for your customers. Probably, your IT landscape is in size not even close to Amazon. Probably comparing to Amazon AWS as the Netherlands compares to the rest of the EU. In most if not all of the dimensions you can think of.

Arguably, the principle of "Automate Everything" doesn't apply to you. I'll leave it up to you to think of one or more arguments why automation is not something you should hold dearly.

Challenge to you: put in the comments a good reason why you think automation is not necessarily needed. I'll make an effort to counter your argumentation as a reply to your comment.
But read on first.

The benefits of automating processes are many. Irrespective of the kind of processes. An automated process is infinitely more likely to be repeatable than any manual process. This results in higher quality since errors will either be made consistently, and can be fixed, or will consistently not occur at all. How compelling is that?
Although the automated and manual version of a process might take the same amount of time to be executed, the automated process allows a person to work concurrently on something else that cannot be automated. And automated processes do not rely on the availability of a specialist to execute the process. So automation makes you and your organisation more scalable.

Not automating processes, even IT processes doesn't make sense. Still, when it comes to IT, we hardly do this. Why?

The situation at Amazon

I've come across many situations where things weren't automated. Worse yet, they could be automated. They were not automatable.

For me this was always an interesting fact to find out. For one, we're in IT and IT is all about automation. In fact, in Dutch we refer to IT as Automation (Automatisering - Dutch). The paradox is that we apply IT all over the enterprise to automate business processes, but when it comes to the IT processes themselves, automation is very likely the last thing on our minds. And when you think of it, that doesn't make sense at all. Walk into a room full of IT people, and just pose the statement that it is hard to understand why we're so good at automating business processes, yet we don't have automation in our own processes. And you'll see at least 90% of in-agreement-nodding heads, the remaining 10% are too flabbergasted with the realisation that this is a true statement. Same statement in a room full of non IT people and the first thing you find yourself doing is explaining why this is.

The fact that IT people are not automating their own processes is tough to explain, and I for one do not have such an explanation. There is an explanation though, for why Amazon AWS's processes are all automated. It's because one of the core principles by which they do IT is "Automate Everything". At a dinner party with Werner Vogels (Amazon's CTO) I was invited to, being the Chief Architect of a FinTech startup, I asked Werner (all attendees were told that we were on first-name basis), how it was possible for a huge company like Amazon AWS to live by these rigorous principles. Everything is an API, Automate Everything, and a few more. His reply was that there were two main reasons why it works. 

1. Senior management all the way up to Jeff Bezos, were openly behind these principles. In fact many of them were mandated by Jeff Bezos himself. 
2. Everybody in the organisation experienced for themselves the validity of these principles.

I asked him to clarify that second reason. According to Werner Vogels, Amazon is dedicating a lot if not all of its time to make its processes impacting customer experience as efficient as possible without impacting customer satisfaction. 'A satisfied customer is a returning customer'. The effects of changing a process is made visible to the whole company, at all times. So compliance to the principles will result in changes to the processes and the effects would be visible. Therefore, the validation of the principles would be continuous. And according to Werner, nothing is as motivating to change your processes than to see the effects of your efforts.

Rest of the dinner I was milling this over, enjoying the food and talk to the other guests.

The situation in the 'real world'

Unfortunately, most of us work at real enterprises. And those two reasons Werner Vogels had given me why IT process automation worked for Amazon aren't obvious in the situations I have found myself in.

For example, the amount of automation in a process within IT is not a metric that anybody is held accountable for anywhere I've worked or consulted. Neither are the benefits of automation part of somebody's accountability.

IT process cost reduction (IPCR) is as far as I know not a KPI within enterprises. Nor is the time-to-market (TTM). The latter often does find itself in another incarnation on reports and dashboards, namely as MTTR, the Mean Time To Resolution. When it comes to MTTR, we do see them on reports, but the MTTR is in hardly any case part of somebody's accountability. Same goes for time-to-market. Although it is always mentioned for any improvement project, it is hardly measured or reported on. It's a project issue in most cases, meaning that an organisation is doing projects instead of delivering products.

The lack of real metrics and if you will KPI's means that from an accountability perspective, there is not a clear person that has an incentive to push automation. And if you read my blogs on a regular basis, you know that I'm big on accountability.
Visibility of the effects of changes to the IT processes is another challenge for organisations. We often find ourselves in organisations that don't have a culture to measure our IT processes. This is especially true for organisations that have been around for decades. Since IT process automation is not pushed, there is no incentive to find bottlenecks in processes or pinpoint areas that are up for improvement. Resulting in the situation where the effect of improvements are not visible in most cases.

The lack of accountability and the rather big hurdle to be taken by IT departments in order to automate result in a situation in which we are just not automating our processes, because it gets no priority on our backlogs or funding in our PRINCE2 budgets. Process automation is collateral. And it's not a matter of not being as large a company as Amazon. It's a matter of not being aware of how automation affects the bottomline. And of not being held accountable for impacting the bottomline.

Third Time Principle

Three times is a charm, or at least should be automated.

In organisations I worked previously, the principle of automation as in play at Amazon was a little bit more pragmatic. One that has worked well for me, is the principle of "Everything done a third time, will be automated", reasoning behind this is that if you do the same thing a third time, it is extremely likely you will do it a 4th, 5th and even more times. The time required to create the automation will be saved by executing the task over and over again.

Time invested is never gained, but can only lead to savings later on.

The Third Time Principle is a compromise, although one might argue that it is a Troyan horse. By adopting this principle, the argument that it creates too much overhead for mundane actions is off the table. Only those actions that are performed repeatedly are automated. Built in justification for the investment needed.

The Third Time Principle is easy to adopt. Since it is a compromise, it can be applied when the push to automation is bottom-up. We often see that engineers see the need for automation since that is where automation will solve a problem and effects are noticed. To develop the automation is often a matter of getting the time to do so. Automation is now competing with other requirements for development time, for priority on the backlog. We all know where the priorities will be. Applying The Third Time Principle will remove this obstacle. Engineers can justify the need for automation and the Product Owner can justify the priority of the automation stories on the backlog.

Continuous Delivery

When striving for Continuous Delivery (CD) and more so for Continuous Deployment (also CD), there is no other way than to automate everything. CD requires a rigorous regime to move every manual task to the left in a process defined western style (i.e. left to right notation). Product development follows the DTAP model. Development is followed by testing is followed by accepting is followed by production → DTAP. In CD we hold true to the paradigm that everything manual is done in D, and TAP are fully automated. Any manual action in T, A and P needs to be automated and the scripts for this are developed in D, because otherwise it can't be done.
So when striving for Continuous Delivery, everything in the product development cycle is to be automated.

Accountability

When process automation metrics are part of somebody's accountability, that person will also be mandated to automate as much as possible. This is for the simple reason that you can't hold somebody accountable for something they cannot influence. Therefore, when process automation is measured through some metric for which somebody is accountable, it is that person's prerogative to drop The Third Time Principle and instead adopt the Automate Everything Principle.

Accountability is a matter of top-down enforcement. It is also a key aspect of culture change. When we want to adopt a culture in which we allow ourselves to reap all benefits of automation, especially our IT processes, we can't get around the fact that we need to revisit the metrics by which we hold ourselves accountable.

Scalability

One aspect that I haven't really touched upon is scalability, although I did mention it briefly. Automation is a key aspect of scalability, not so much scalability on a technical level but organisational scalability. Manual actions always need a person to execute them. The more complex the activities become, the more experienced or knowledgeable the person needs to be. And before you know it, it requires a specific person within the organisation. Because you rely on this person, that person will become the person with all required knowledge, it's a vicious circle. Think about it for a second, and I'm sure that you can think of a process or an activity in a process where you rely on a specific person and you know that person by name. In fact, if you want the activity to be done, you will even call that person because she's among the busiest persons in the company.

When you want to keep activities simple, you need to automate them. The more you automate, the easier it is to keep the activity simple. Hardly anything is more simpler than to have a command like 'do_complex_activity.py' provided that the automation is done using Python. Anybody can run this command provided they have the rights to do so, and when done really properly, anybody can do it at any time, because all the complexity of who can do what when is taken care of by the automation code as well.

Concluding

In the end, you'll agree that automation is what we need to apply to all our processes. At least when it's the third time you're doing the same job. It will improve quality and consistency. It also means that we as a company can scale our organisation without the need to increase headcount. It will be important to monitor the benefits of automation, without metrics it will remain a matter of good faith on the short term, and it will not justify the investments on a longer term. By understanding the benefits of automation and getting insights on where automation has most impact.

Thanks once again for reading my blog. Please don't be reluctant to Tweet about it, put a link on Facebook or recommend this blog to your network on LinkedIn. Heck, send the link of my blog to all your Whatsapp friends and everybody in your contact-list. But if you really want to show your appreciation, drop a comment with your opinion on the topic, your experiences or anything else that is relevant.

Arc-E-Tect


The text very explicitly communicates my own personal views, experiences and practices. Any similarities with the views, experiences and practices of any of my previous or current clients, customers or employers are strictly coincidental. This post is therefore my own, and I am the sole author of it and am the sole copyright holder of it.

The "Eat your own dog-food" fallacy

Why having people eat their own dog-food doesn't accomplish anything sustainable.

Summarising

Having somebody eat their own dog-food doesn't necessarily make the improve its taste but might make them get used to the taste instead. Awareness of the (lack off) quality in one's work is important when transitioning from a Dev/Ops organisation towards a DevOps organisation. But experiencing a lack of quality first hand is often not the way to do so, or even an option. Agile transformations are only possible by means of cooperation.

There's a common understanding in the world of organisations that are transforming from Dev/Ops towards DevOps organisation that this works best when the devs are forced to eat their own dog-food. It's a reaction from the Ops people towards the Dev people.

Basically it means that once you've got the Devs supporting their own products, they'll make sure that those products are of a high quality. Common believe is that since they, those Devs, don't want to get called Sunday night at 3 AM because their software crashed. Which makes perfect sense, who does want to get called Sunday night at 3 AM because something they created crashed? I wouldn't, would you?

So, if you want those Devs to focus more on quality, on less crashing software, you have to make them support that same software themselves. In other words, turn those Devs into Ops people and that'll show them.

About a year ago, I joined a team at one of my customers to help them transform selected development teams into more agile teams utilising Continuous Delivery mechanics and move towards, lord forbid, DevOps. One of the slogans we used to get the necessary buy-in was:

"Eat Your Own Dog-Food"

And later we added "And Clean Your Own Shit!". Totally convinced that this is how it works. Make people feel the pain they're causing and they'll become better persons. If you would do a little time-travel and rewind about 2 years, you'll hear me saying that "one should feel the pain they're inflicting".

I think you'll recognise this when you've been in that situation where you want to move from Dev/Ops towards DevOps. Or become more agile. Or maybe you need to convince people that agile or DevOps is the way to go.

It makes total sense. People with kids know this. You want them to stay away from the fire, let them burn their fingers. Or those people with dogs shitting all over the place... let them clean that shit every time their dogs drop their poop in the playground. It works, really does.

But there's a huge problem in this, I'll get to that. First I want to ask you if you've noticed that slightly grumpy undertone in me mentioning of the "Eat your own dog-food" slogan and everything associated with it.

Agile and SCRUM in particular are developer driven ways of working. It's the developers that want to change things. Reason, obviously, is that developers want to develop software that people are using, so they want to get create something that is as close to what is usable as possible. Meaning that they want so put something in the hands of the user as quickly as possible and then make adjustments and put that into those same hands. And again. And again. And again.
Our organisations are such that specialised Ops people are managing those applications when in the hands of those users. And they need to keep up with all those adjustments... You understand the predicament those Ops people are in.
So when you tell these Ops people, that you want on your side while transforming into agile organisations that those Devs should eat their own dog-food. Ever tasted dog-food? So how do you think that this resonates with those Ops people? Pretty awesome, don't you think?
That connotation of dog-food is getting the nay-sayers called Ops on your hand, shouting "Yay" instead of "Nay". Mission accomplished. You're done.

Well forget it. That's the "Eat your own dog-food" fallacy. It's a fallacy because it doesn't solve anything and it certainly won't help you in your agile transformations. Considering your organisation has separate Dev and Ops teams, and considering that the reason for this is a more efficient Ops team because they will support many products. Because supporting a quality product is not a full-time job. Read my post on this topic. There's no way that having the Devs eat their own dog food will improve quality. Which was the premise in the first place, remember? And now you should say that in fact it doesn't make sense at all. Because there's an Ops team and for a good reason. So what makes you think that anybody in the organisation will just allow the Devs take over the Ops jobs? Ain't gonna happen. No way. Meaning that whatever you're trying, the Devs won't even get the opportunity to eat their own dog-food, even if they would want to. You can fix it by job-rotation... in certain situations, in certain organisations. Fallacy explained.

Considering the above, how would you need to address the agile transformation? How to move towards a DevOps setting? The answer is quite simple and probably extremely hard to implement. The more alluring the "eat your own dog-food" approach is, the harder it will be to do the correct thing. The sustainable approach, let's call it.

If you want the developers to develop software of a higher quality, you need to make them aware of the problems they are causing because of the lack of quality in the software. And you do this, by introducing them to their operations colleagues. Let them work closely together. Geographically close. As in side by side. Not pair programming close, but across the desk close.
What will happen is that the colleague from Ops will complain to his Dev colleague about a problem instead of to his Ops colleague. Feedback loop is tiny and closed. And most likely, it's friendly constructive feedback, because it's directed to the immediate colleague from across the desk. That person with whom lunch is shared. Not bottled up feedback. It becomes a feedback loop in which the Dev can immediately ask the Ops person why it's such a huge problem, this tiny thingy.
Getting the Dev and the Ops to work at the same desk, allows them to become aware of each other's work. And generates understanding. Understanding towards their respective worlds. It creates an atmosphere where the Dev will improve the quality of the product, because otherwise the Ops colleague is called Sunday morning 3 AM, and that's not something the Dev wants.

As a parting gift, another tip: Make sure that Devs and Ops don't huddle together when you put them in the same room. Instead put the Dev next to the Ops, side by side, hand in hand. When you allow them to huddle together, you should put them in separate rooms. Just to make sure that their respective complaining is not effecting them during work hours.

By allowing those to blood-types in your organisation to become aware of each other and befriend each other, you'll pave the way to become a true agile organisation with a smooth transition towards a DevOps mindset.

Here's an exercise for you: See how it works the other way around. Feel free to use the comments to discuss.

Thanks once again for reading my blog. Please don't be reluctant to Tweet about it, put a link on Facebook or recommend this blog to your network on LinkedIn. Heck, send the link of my blog to all your Whatsapp friends and everybody in your contact-list. But if you really want to show your appreciation, drop a comment with your opinion on the topic, your experiences or anything else that is relevant.

Arc-E-Tect

Microservices on steroids, getting from just agile to business agile [part 2 of 2]

Summary

This is how you get Microservices on steroids; apply them in a Policy Oriented Architecture. Meaning that you need to put a policy handler in between a service's interface and its implementations. The reason why you want to do this, is because you want to provide agility to your business and not just have an IT department that's agile. Or in fact have an IT department at all for that matter.

It’s time for part 2 of a two part series on Microservices on steroids. I already had part of the post ready but today I was in a presentation on the topic of Continuous Delivery and Architecture and one of the aspects that came up was about ‘Product definition’ and I figured that I needed to put that into the mix as well.

The conclusion of the previous post (Microservices on steroids [1/2]) was that adopting a Microservices based architecture, which pretty much is an SOA the way it was meant to be, will only get you technical agility. As you can read in my post on Continuous Delivery not being something for the IT department only (The Continuous Delivery IT Team Fallacy), technical agility is only part of the story and in fact has limited value if you’re not targeting business agility.

But let’s first consider what Business Agility exactly means. It is in fact exactly what it literally means, agility of the business. The main difference with the ‘normal’ agility is that it allows the business to change its course rapidly without a direct need of involving IT to make this change. So without needing to worry about whether or not IT can handle these changes.

And it allows the business to respond to changes in its marketplace when and where necessary with the help of IT. So business agility is an extension of technical or IT agility.

An example of business agility is for example bank XYZ that issues debit cards to its customers age 16 and up. One of its competitors, bank ABC, is in a marketing campaign targeting teenagers and issues debit cards to 12 to 16 year olds when a letter of consent is signed by their parents or legal guardians. Our bank XYZ will lose out on a large customer segment when not also addressing these teenagers. So it will need to issue cards to them as well. Business agility is when bank XYZ can do so, without major activities that need to take place in order to change course. Think about being able to make this business change in 1 two-week sprint.

From experience I can say that an agile business requires an agile IT that has its architecture on track. An SOA, preferably based on Microservices, is the best basis for this. If you ask me, I would stay away from centralized components like single ESB installations for an organization, shared database clusters or API management solutions that are centrally governed. It’s not so much a matter of me disliking these technologies, which I really don’t. It’s about the centralized governance issue, which results in shared resources, which seriously hampers independence. Centralization also means, artificial, restrictions in autonomy. In effect, it means that you reduce your ability for agility.

Now, don’t get me wrong, centralization doesn’t have to be a bad thing. In general it allows you to limit your costs by benefitting economies of scale. It’s a way of improving profitability by reducing costs. There are many situations in which this is the best way to address profitability. Especially in areas where business functionality is established and there is not really a need any more to figure out your product-market-fit, focusing on cost reduction is good. Within the same organization you also want to be able to improve profitability by increasing revenue. This is especially true for those situations or products for example where product-market-fit is a challenge or when the business scope of a product is changing or increasing still.

 

Think back a couple of posts where I explained my stance on Gartner’s Bi-model IT (The BI-Modal Misconception...). The two situations I mention are the closest thing that comes to Bi-model IT that actually makes sense. It’s where the (business) need for change and agility has diminished and functionality is stable vs. where the need for change and agility is absolutely there and business survival depends on it. There’s not a single thing that relates to risk or quality or ability. It’s about the need to be agile or not that defines the need to become agile or not. And typically, Mode 2 is revenue increase focused where Mode 1 is costs focused. Gartner’s emphasis on legacy transforming to the digital world is not relevant at all either.

Back to business agility. This is where the business is agile as established before. And like I stated earlier, this requires an architecture in which the different components are independent and can be independently deployed. Where ‘components’ are ‘products’. Independently deployable components are almost the same as Microservices. And although Microservices is a buzz word, it does have significant merit to base your architecture on Microservices in order to deliver agility to your business. It’s the perfect foundation for business agility.

Which gets me to the point of my presentation on the topic of Continuous Delivery and Architecture I mentioned earlier.

It is a common misconception that Agile and Architecture don’t play well together. Which is over course utter BS, and if you don’t believe me, I posted about this not that long ago (Product Owner and Architect, Agile Tag Team). Agile and Architecture play extremely well together as long as the Product Owner and the Architect(s) play well together. Especially when it comes to Microservices you need the domain architect or the business architect or both when you have that luxury. It is the architect’s responsibility to understand the business domain and together with the PO define what products make up that domain. Every product is a very likely candidate of either become a Microservice in its own right or becomes a composition (or constellation) of Microservices. And the Product Architect, which you might know as being the Application Architect, Project Architect or Solution Architect, is the single one person that together with the PO defines where the Product ends and the Platform on which it runs starts. The Product Architect therefore is the person that defines the (technical) boundaries of the Product and the domain architect defines the product’s place in the IT landscape. And like I said, a product is a Microservice or constellation of Microservices. Ideally, of course. Domain architect and Product architect should also work closely together in defining what API’s to consume and API’s to provide. Again together with the PO since it’s all about the PO’s product.

In this situation, we have a nice decomposition of a product or several products in interfaces and implementations. And that is exactly what we want from an SOA, especially one that is comprised of independently deployable components, services if you will. Or even Microservices.

Once all interactions between products and within products are based on interfaces, we can talk about a true SOA.

Now here’s an interesting detail. Interfaces are nothing but documentation. Be it fairly intelligent documentation or rather very usable documentation, but documentation nonetheless. And everybody that thinks an interface is something else is wrong. We don’t put an interface in an architectural layer because that doesn’t make sense. The same goes for API’s. An API is also nothing more than documentation that conforms to specific standards. Of course it’s a bit more than just a document, but really, just a little bit more. An API without an implementation is nothing and in fact you can’t actually deploy an API, nor can you an interface for that matter.

Having established that, it’s time to start the confusion and get on with it big time.

So, since an interface is nothing but documentation and it’s something you can’t really deploy. In order to provide business agility through an SOA, we need to put something in between interfaces and their implementations… But before I venture into that area, let me explain how you get business agility through an SOA.

We do this by not implementing an SOA but a POA. Where an SOA is a Service Oriented Architecture, a POA is a Policy Oriented Architecture. Basically this is an architecture in which you have services separated into an interface and one or more implementations of the interface. All implementations are available concurrently and based on policies, one of the implementations is selected. The policies are business policies and governed by the business. Business policies are a fancy word for business rules.

An example of such a business rule is that when transferring money between a customer’s own accounts there is no need for additional validation of the transaction once the customer is signed into her internet banking. Another rule is that when transferring money from a customer’s own account to an account of another customer, an additional signature is needed. Both are financial transactions, but based on the context, different business rules apply.

Policies are context driven, which amounts to the notion of under what circumstances a service is consumed as that determines the implementation of the interface to be used. Contexts can be everything imaginable. Within the example of the transaction as mentioned above, it can be the amount of the transaction, the age of the customer, the time of day, the geographical location of either customer in the transaction, the type of accounts, the client device used and the way a customer was authenticated. Or any combination of these contextual parameters or something completely different altogether.

In a POA, there is a clear distinction between interfaces, policies and implementations. The policy sits between the interface and the implementation. And it is business definable, meaning that the business rules can be changed without changing the interface and existing implementations. Either resulting in implementing policy compliance, i.e. some software is developed that implements the functionality to comply with the policy or by implementing another implementation of the service.

Some 10 years ago, Aspect Oriented Programming (AOP) was hot in software development country and although this way of software development is more or less forgotten, it is exactly what POA (see the similarity in acronyms?) wants to establish. By injecting new code, or rather new aspects of an implementation policies can be changed.

The problem with AOP was that was not trivial to do and it was hard to debug once there was an error in the functionality. AOP is also at the class level instead of the service level, and it is far from accessible to non-developers. Still the paradigms of AOP are what POA intends to deliver. A more flexible way of changing nuances in an interface’s implementation.

As you can understand is that the role of the architect and the PO is very significant in POA. The Product Owner needs to understand where policies are to be in place from a business perspective. Not every interface needs policies, and sometimes you don’t even want to have the ability to change a policy without strict governance. The architect, and typically the Product Architect, needs to define how policies are implemented. Since there are as many ways of implementing policies as there are use cases for policies, it is important to at least within the context of a product the same mechanism is used. This is where the architect comes in.

The power of an API, or a Microservice or even a normal service, is so much more when POA aspects are applied as the Product is enabling or rather facilitating business agility. Especially when the used framework for policy definition and implementation is one that allows for doing so by a non-developer. Unfortunately there are not that many frameworks available, most solutions are based on Case Management solutions or consist of glorified Business Process or Business Rules engines. Centralized solutions that require specific expertise to operate and maintain them. Consequently, the autonomity and therefore the agility of the teams and business users are limited. This is where this post more or less started off from. Centralized solutions are limiting agility, and when you want to extend agility to the business, you should limit it in the technology.

So this is how you get Microservices on steroids; apply them in a Policy Oriented Architecture. I hope you liked this post and feel free to comment and discuss. Although the POA is not new, you don’t see it applied that often if at all. I don’t think it’s a matter of complexity or a technology issue. The challenge is typically in that business and IT need to be considered as one and not two separate departments.

Thanks once again for reading my blog. Please don't be reluctant to Tweet about it, put a link on Facebook or recommend this blog to your network on LinkedIn. Heck, send the link of my blog to all your Whatsapp friends and everybody in your contactlist. But if you really want to show your appreciation, drop a comment with your opinion on the topic, your experiences or anything else that is relevant.

Arc-E-Tect

‘The Continuous Delivery IT Team’ fallacy

Summarizing

Considering Continuous Delivery something for your IT department is throwing your money out the window when doing an Agile Transformation program. If you want to do so, make sure you throw it into my direction. IT is a business concern, Continuous Delivery is a business concern.

Over the past couple of months I did a series of awareness sessions on Agile, Continuous Delivery and DevOps at a large client of mine. As is rather common, also at this organization the initiative to move towards Agile ways of working, Continuous Delivery and a longing for DevOps is with the software developers. This makes perfect sense when you think of it, because it's the developers that want or rather need to make changes and the pace by which they have to deliver changes is only increasing. But I guess I don't need to tell you this.

But Continuous Delivery (or Deployment for that matter) is only possible when you don't consider it a software development thing. There really is no point in spending any time to move to Continuous Delivery when you are not planning to do it broadly. I'll get to that in a second.

During these awareness sessions, which I do with colleagues from the same team, we discuss with non-developers like risk officers, sys admins, project managers, architects, test consultants, etc. we outline what Continuous Delivery and DevOps mean within the context of my client's organization. It's a common story and I won't bore you with the details, but obviously we touch upon the benefits of small increments, feedback loops and so on. And the fact of the matter is that our story actually does sound like it's the perfect thing. And we consistently get the question: "So, really, it can't be all awesomeness, so what's the catch?". And in the rare occasions we don't get this question, we still answer it.

The biggest problem with Continuous Delivery and everything following from it, is that it is not a software development thing, or even an IT thing. When you think so and still go down that road, spare yourself the frustration and disappointment, drop me an email asking for my IBAN number and transfer the budget for your transformation project into my account. At least one person will be happy with you spending that money.

And yes, even when you think it's an IT thing, you're better of giving me that money. This is what I call:

'The Continuous Delivery IT Team' fallacy

Let me explain. First of all, let's make sure you understand what I consider Continuous Delivery. It's the process that produces a product that results in business value in order to be able to sustain the organization up to the point that it can be delivered to a user and it will be delivered to a user as soon as possible. It's not a perfect definition or even a formal definition, much because there's so much more to it. But what's important is that it defines work on a product as being done, when it is ready to be delivered to a user. The actual delivery is an explicit manual step which is decided by the Product Owner to be taken as where the rest of the process is preferably fully automated. This as opposed to Continuous Deployment, where the delivery to the user is also automated and hence triggered by the developer when committing his code to the source code repository. Again, this is a definition close enough to what it is and fitting the purpose of this post.

With Continuous Delivery and agile working in general you want to receive feedback for what you've been doing as soon as possible. And preferably you want this feedback to be such that for one you know that you've done well and secondly you're actually contributed to the bottom line of the organization you work for. This is why we like to work at the granularity of user stories and epics and delivery on a per user story or at least on a per epic the changes to a user.

As you now understand, for every single user story or at least epic, you need to do everything that needs to be done for a release, because you're delivering to a user. Somebody is going to actually use that little piece of software you've worked on with such a passion. Fact may be that the complexity is limited because increments are small, still you need to release new or changed software. And there's the catch.

With software delivery, or product delivery in general, it's not just the product development team that's involved, or more specifically the software development team. It's other teams and people involved as well. Think about marketing, legal and compliance, worker's associations, security and risk management. These are all teams that are not part of the IT department. And no, security officers are not part of an IT department, and in case they are, they most definitely shouldn't be. And the biggest catch of all, the 'business' needs to be involved from day -1. Unless all of these different roles, teams, people, stakeholders, however you want to call them are on board and work in those same small increments, not becoming a bottleneck and automate as much as possible, your Continuous Delivery efforts are a waste of everybody's time and your organization's money.

Back to your 'business', it's them that request for features and not the user. It's them that pay for the development of the product not using it per se. When that 'business' is not capable or willing to define the product's features such that they can be delivered in tiny chunks, than you're out of luck and not much will come from Continuous Delivery in your organization.

The Product Owner is key in all of this. Being the hinge between the Product Team and the rest of the world, the PO is the single one person that can and must ensure that the product is delivered incrementally, with business value visibly added with each increment. PO can't do this, you've got yourself some trouble. The PO, at all times, must be able to relate every single feature, one way or another, to an improved life for the user of your product and therefore a positive effect on the organization's bottom line.

So, Continuous Delivery is not an IT thing, it's a business thing. And don't let anybody convince you otherwise. Being as it is, moving towards an Agile way of working and Continuous Delivery or Deployment, in fact would mean that you no longer consider your IT as being delivered by a separate IT department, but as an integral part of your business.

This does make perfect sense, considering your IT as part of your business I mean. It makes perfect sense because more and more business are all about information and are run based on information. IT is no longer a tool, like a glorified typewriter, it is in fact what is producing business value. No IT no business.

Thanks once again for reading my blog. Please don't be reluctant to Tweet about it, put a link on Facebook or recommend this blog to your network on LinkedIn. Heck, send the link of my blog to all your Whatsapp friends and everybody in your contactlist. But if you really want to show your appreciation, drop a comment with your opinion on the topic, your experiences or anything else that is relevant.

Arc-E-Tect

Microservices on steroids, getting from just agile to business agile [part 1 of 2]

Summary

Agility is important, and business agility even more important. With this, the architect has to play a crucial role by defining the right principles, commandments if you will, and seeing to it that the product development teams live by them. Principles like KISS and Independent Deployability are two principles not to mess with. MIcroservices are an important and extremely helpful aspect of this, so is continuous delivery. But you need to get those microservices on steroids in order to be able to move from just agile to business agile. In this first of a two piece post, I am laying the ground works for explaining how you can facilitate business agility in your architecture.

Since you’re reading this post, it’s likely you’re an IT architect or have to deal with one every once in a while. Meaning that you’ve been confronted one way or the other with Service Oriented Architectures (SOA). Maybe you’re even a proponent of this architectural style.
I’m not going to explain what an SOA is, there’s quite a few very good sources available online explaining what it is and why it’s a good way of building your IT landscape, or not. As always, there are just as many proponents as there are people against and SOA. Over the years I’ve grown to like the premise of an IT landscape where fairly small pieces of functionality can be strung together to realize business functions. And with the advent of RESTful web-services, the premise of a true SOA has become more prevalent as ever.

Nowadays all you see and hear around you are Microservices, even though there doesn't seem to be a clear definition that everybody feels comfortable with, everybody is doing Microservices. One of the most interesting definitions I heard is that it is a really small service. But not as small as a nano service. But just to make sure that you know what I refer to when discussing Microservices, here're two definitions I like to use. The first is from Wikipedia, which in and by itself is no guarantee for it to be correct, but at least it's a common definition:

"Microservices is a variant of the service-oriented architecture (SOA) architectural style that structures an application as a collection of loosely coupled services. In a microservices architecture, services should be fine-grained and the protocols should be lightweight. The benefit of decomposing an application into different smaller services is that it improves modularity and makes the application easier to understand, develop and test. It also parallelizes development by enabling small autonomous teams to develop, deploy and scale their respective services independently. It also allows the architecture of an individual service to emerge through continuous refactoring. Microservices-based architectures enable continuous delivery and deployment."

The second definition is form Martin Fowler, which is in essence the same:

"In short, the microservice architectural style [1] is an approach to developing a single application as a suite of small services, each running in its own process and communicating with lightweight mechanisms, often an HTTP resource API. These services are built around business capabilities and independently deployable by fully automated deployment machinery. There is a bare minimum of centralized management of these services, which may be written in different programming languages and use different data storage technologies."

Microservices are not easy to develop, much like API's, they require a rather typical level of maturity of a development team to develop them. The hard part of Microservices is typically the "independently deployable" aspect of the services and then in particular maintaining that independence of time. Let's call it sustainable independence.

What I usually see is that over time the architecture becomes convoluted, resources entangled (the whole HATEOAS thingy is not trivial at all) and adversity against rearchitecting the IT landscape to get back to independent microservices.

To develop them,  the Microservices, you need to invest. In general, a Microservice is way more expensive to develop than a regular service. Much like an API being a lot more expensive than a simple service interface. So I would argue that there needs to be a business case to capture some functionality or resource in a Microservice.

Still, Microservices are a good thing to look into, if not only to be able to dismiss them. But more importantly, they are very helpful in becoming agile. I would almost argue that in order to be agile, you need an SOA based on Microservices and API's. Having said this, I need to emphasise that we're still talking about agility of the product development team, the IT people.

Mind that having an architecture with highly independent units of deployment is a blessing for all of us that want to do Continuous Delivery or Deployment for that matter. It makes life so much easier and as an architect you should really make sure that independence of components, services or units of deployments, however you want to call it, is a prime-directive to all your developers, great and small. It should be up there on the wall with the rest of your architecture principles. First of course you mention KISS and second the declaration of independence for your components.

But as I said, agility so far is technical agility. You've designed your software, your product, for change. Which allows you to be agile. Your PO (Product Owner) wants something different, you can change your code easy as can be. But when you think about it, it won't make your business more agile, it will make your IT more agile.

There's quite a bit more to be done before your business gets its agility... more on that in the next part.

Thanks once again for reading my blog. Please don't be reluctant to Tweet about it, put a link on Facebook or recommend this blog to your network on LinkedIn. Heck, send the link of my blog to all your Whatsapp friends and everybody in your contactlist. But if you really want to show your appreciation, drop a comment with your opinion on the topic, your experiences or anything else that is relevant.

Arc-E-Tect

Special thanks to Sytse for pointing out one or two textual errors and helping me correcting them.