- “Gartner about Microservices – is this a Big Mistake after WebServices?” https://www.linkedin.com/pulse/gartner-microservices-big-after-michael
- “Application Architecture for Digital Business” http://my.gartner.com/portal/server.pt?open=512&objID=202&mode=2&PageID=5553&ref=webinar-rss&resId=2866521&srcId=1-3134301641
- “Microservices vs SOA - Is there any real difference?” https://genehughson.wordpress.com/2014/12/17/microservices-vs-soa-is-there-any-real-difference/
- “µCon 2014: The Microservices Conference” https://skillsmatter.com/conferences/6312-mucon
- “Can I Haz Name?” http://203.softover.com/2014/12/01/ultra-srevices-and-declaration-of-independence/
- “#BPM for software architects - from monolith applications to explicit and executable #coordination of #microservices architecture” http://improving-bpm-systems.blogspot.ch/2014/08/bpm-for-software-architects-from.html
- “#BPM for the #digital age – Shifting architecture focus from the thing to how the things change together” http://improving-bpm-systems.blogspot.ch/2014/08/bpm-for-digital-age-shifting.html
- “e-government reference model #GeGF2014 #egov #entarch #bpm #soa” http://improving-bpm-systems.blogspot.ch/2014/10/e-government-reference-model-gegf2014.html
- “#BPM for #SOA+#ESB+#API and #cloud (#PaaS and #SaaS)” http://improving-bpm-systems.blogspot.ch/2014/12/bpm-for-soaesbapi-and-cloud-paas-and.html
- "Ideas for #BPMshift – Delenda est “vendor-centric #BPM” – How to modernise a legacy ERP" http://improving-bpm-systems.blogspot.ch/2014/04/ideas-for-bpmshift-delenda-est-vendor_27.html
And other references:
At the same time, it is considered that there is no commonly-agreed definition of “microservice”. One of the reference definition of “service” is in http://www.infoq.com/articles/updated-soa-principles .
My definition of “service” is “explicitly-defined and operationally-independent unit of functionality”.
Application architecture current challenge
the microservice architectural style ...is an approach to developing a single application as a suite of small services.
In other words, the question is how to implement an application as an organised collection of many autonomous components. Each Autonomous Component (AC) is (potentially) distributed (in-house or in-cloud) and thus it is a unit of deployment (otherwise it can be deployed in a separate host).
Note: At present, it is not possible to claim that ACs are microservices or services.
I think that this is a natural step in the evolution of the application architecture.
From monolith: one unit of deployment, one “in-process” (in techie’s meaning – in the same JVM), simple inter-component communication (again “in-process”) and simple error-handling (again thanks to “in-process”).
Then moved to client-server with two units of deployment (but not yet autonomous components): fat client (presentation and some biz logic) and the rest.
Then to three-tier with three almost ACs: thin client (presentation, better UX), business logic (calculations) and data access layer.
Now, with the recent popularity of mobile equipment, in addition to IT-centric decomposition (presentation, logic, data) into ACs, there is an opportunity for functional decomposition into ACs. So, each AC may have its own UI/presentation, logic, data and additional access channels, e.g. classic API. An example of functional decomposition is a portal – a data-centric portal for navigation over some data or a function-centric portal or a combination of them.
Meaning of “autonomous”
The word “autonomous”, being a keyword, requires an extra explanation. Some people associate it with “...Independent Scalability, Independent Lifecycle and Independent Data” but, let us look at the full life-cycle of AC.
In general, such a detailed life-cycle has the following phases:
- Contextualise (or define future AC’s usage/expenses - WHY)
- Plan (or schedule AC’s design, build and run - WHEN)
- Design (or define AC’s characteristics - WHAT)
- Build (or implement AC -HOW)
- Bind (or linking)
- Deploy (WHERE)
- Monitor (or meter of the usage of AC - WHO)
- Measure (or evaluate the performance of AC – WITH WHAT RESULTS)
Considering that each AC will participate in one or many assemblies for provisioning a “bigger” solution (or richer functionality), ACs are interdependent at their “creation” phases (contextualise, plan, design, build). Imagine that a highly-efficient team is created by the careful selection and training of people.
Ideally, ACs are independent at their “operation” phases (deploy, run, monitor and measure). For example, a high consumption of CPU by one AC will have no negative effect on other ACs. Or, an AC may be stop, un-deployed, re-deployed and resumed without determination of the whole assembly performance.
But, the binding or linking of ACs into an assembly makes ACs dependent in accordance with their contracts. Up-steam dependency is related to the AC’s performance. Down-stream dependency may happen as well by imposing some ACs to be used. (Note: There are several techniques to reduce dependency).
Thus, autonomicity of each component will vary from more to less (independency, dependency and interdependency) at the different phases of the full life-cycle.
Structurally interdependent, behaviourally independent and contractually dependent (Note: there are behavioural rules like in a red-card football) – what a mixture.
Working together in assembles
There must be a set of common rules for AC to help ACs work synergistically together (as an assembly) for a common goal. I think about the following rules:
- each AC acts similar to services,
- AC may have a particular specialisation: user-facing (interactive service), coordinator (orchestration, etc.), lawyer (decisional business rules), utility (basic functionality), resource (data, etc.), communicator (lightweight moderator), dispatcher (event handler?), referee (behaviour rules enforcer), porter (security service), registrant (naming services), etc.
- to work together, ACs follows agreed admission rules: naming, formally defined interface, performance, behaviour, etc. Admission rules for different assembles may be different.
- in particular, ACs may have to agree using some common ACs with assembly-wide functionality.
In extreme, ACs may work together as a group of fully-universal-agents (no specialisation) or as a team of specialised ACs (some team-sports). Particular type of specialisation may be a coordination of ACs with “small or simple” functionality into an AC with “big or complex” functionality. (For example, combining “generate PDF”, “protect PDF by a digital signature” and “disseminate PDF” together.)
Let us try to use ACs for some modern ideas about a better application architecture (again don’t be fussy with the terminology) as one of them cites below:
“A modern application is a functional ecosystem comprising a loose association of apps and services. Apps implement the application front end (user-facing) to support a specific functionality on a particular type of device and interaction medium, and services implement the application back end. Together these apps and services support a particular business domain.” (Note, also RIA were talking about this).
I think that the following types of AC are necessary to cover the majority of “modern applications”:
- various resources (primarily, data) – this is a typical data-access AC.
- utilities – this is a typical data/documents transformation AC.
- data/documents role-based mini-portal (select data/documents and initiate an operation on them) - this is a user-facing AC.
- functional portals (select a function and execute it on some data/document) – this is a user-facing AC.
- short-running (or synchronously executed or near-immediate completion) operation – this is a user-facing AC with some logic and invocation of some other ACs.
- long-running (or asynchronously executed) operation – this is a orchestration-like AC for long-running coordination of people and ACs.
- stateless and idempotent composition of some ACs for a composite operation – this is a low-code aggregation for invocations of other ACs.
Archetypes of application which can be constructed from ACs
Data-centric application archetypeTypical usage: browse a data repository, select and execute an operation.
Document-centric application archetypeTypical usage: browse a document repository, select and execute an operation.
Example: document management systems
Operation-centric application (with short-running operations) archetype
Typical usage: browse a static list of operations, select an operation, associate with it some data/documents, execute this operation.
Example: employee portal
Note: all operations are short-running – they are completed in less than a few minutes (so a person can wait for their completion in front of the screen).
The static list of operation is role-dependent.
Operation-centric application (with long-running operations) archetype
Typical usage: browse the static and dynamic lists of operations, select an operation, associate with it some data/documents, execute this operation.
Example: employee portal.
Note: all operations are long-running – they are completed in a few days / weeks (so a person cannot wait for their completion in front of the screen).
The static and dynamic lists of operation are role-dependent.
Real applicationsThere may be a mixture of mentioned above archetypes.