2. About Mike…
• Software Engineering Manager at CaringBridge
• Open Source Contributor
• Organizer of MNPHP
• Where you can find me:
Twitter: mwillbanks
G+: Mike Willbanks
IRC (freenode): lubs
Blog: http://blog.digitalstruct.com
2
4. Definitions
• SOA – Service-Oriented Architecture
A set of principles and methodologies for designing and developing
software.
A structure for delivering content through services.
• Think of a service consumer attaching to your web service; it likely is
not using a database.
• Service
A set of related software functionalities that can be reused for
different purposes.
Essentially; the encompassed business functionality divided into
tasks that streamline the interface and remove the spaghetti.
4
5. A Question to Consider
• How often have you thought about not only how your code
can serve others but how it can service you, your team
along side of supporting external parties?
Think core site.
Think mobile.
Think public.
5
6. What I mean by that…
• Slight adjustments when not using a service based model
has a larger scale of changes throughout your app.
• Adding caching or logging can be more difficult.
• Your entity was split into multiple tables but you utilized a
Zend_Db_Table for the model class. It is going to be a
larger change.
• Business Process in the Controller; forces implementation
into the additional products and if it changes has down
stream expense.
6
8. Zend Framework and SOA
Service Layer
Decorating Models
Attaching the Servers
9. Service Layer
One of the most important things when building a SOA layer
is going to be the underlying services; these services are
enabled by the architecture of your models.
11. A Service
• Satisfies a Business function
Function broken into Tasks
• Easy consumption by your application
• Allows vast re-use and more generic implementations
• Decoupled to represent more or less a single item
• Application logic becomes more encompassed
Filtering
Validation
Transactions / Interactions between Domain Models
11
13. Domain Models
• A domain model is a representative entity of “something”
that solves a domain problem.
• The entity is NOT explicitly tied to your database.
• The entity essentially contains getters and setters for
properties and potentially some behaviors.
13
15. Data Mapper
• A data mapper handles the communication between the
data source and the population of an domain model.
In a ZF context this is generally fetching data from
Zend_Db_Table.
• Extremely flexible and easily handles if the underlying data
changes.
Alternatively; when you now have separate tables that still can
represent the same domain model.
15
17. Data Store
• Persistence… you need to store it somewhere!
• Data may come from several places…
File System
Web Service
NoSQL DB
Relational DB
• If you utilize an ORM this is also where it would live.
17
20. Decorators?
• Let’s face it… decorators do a better job decorating than we
do ourselves.
20
21. Decorators
• Decorators allow you to add new behaviors to an existing
object dynamically…
Remember the previous examples? We used a few interfaces and
now you will see why.
• General Use Cases
Performance is becoming more of a concern and you need to
implement caching.
An object is coming back with unexpected data at times; so you
need to be able to log what was coming through.
You need to represent an object differently such as JSON, XML,
Serialized PHP, etc.
21
25. Just Decoration?
• Remember decoration is just a pattern; it can help you do
several things but it cannot do everything!
• Don’t be afraid to utilize other patterns to assist you in the
build out of your models.
• The ultimate goal here is to remove the constant cost of
refactoring for every feature change.
25
27. The Server Components
• All of the server components act approximately the same in
terms of their API.
• Each of them follow the same paradigm allowing you to
connect them directly to a service object.
That means… this is going to be easy!
• However, the server components utilize Reflection for
AutoDiscovery as well as to understand the actual call.
27
29. Implementing Servers through Controllers
• Why use controllers?
Because they are already there and it fits directly into how you
made your controller in the first place.
You can leverage your existing code to add them in.
Providing only a new action makes this possible.
Easily discoverable just by the URI.
• In our examples – we are going to leverage an API action
29
31. It really is just that “easy”
• We just implemented a JSON RPC server, wasn’t that easy?
• You now have methods that you can call through a JSON
RPC client based right off of your service object.
• We did skip out on something, isn’t there a discoverability
component?
31
33. The Community Wants More
• The API is now exposed, people are using it but you are
getting requests for additional protocols
Where is the SOAP API?
How come there is no XML RPC API?
Everyone uses more REST based API’s, WTF?
• Time to implement some more!
For now, we will implement this using a type variable.
33
35. So What About those Formatters?
• Previously in our decoration; I showed you how to change
the output by decoration.
• While they can be useful sometimes; Context Switching
handles much of this for you already.
So it was really just an example but may become relevant
elsewhere.
35
38. Versioning Services
• There are a few different ways to version services; we know
things will change and we need a way to support BC.
• Versioning is important and can be implemented several
ways.
• They are not “perfect” but can and will work.
38
39. Versioning by Directory
• Make version directories in your models directory
Pros
• Quick to implement
• Individual services take on a new version
Cons
• Not all services take on the same version number; more confusing to
the consumer of the service.
• You need to push a version parameter to your api action
• You would end up implementing fallbacks based on versions not
existing.
39
40. Versioning by Module
• Make version modules
Pros
• Separation of entire service into a new module
• All services take on a new version.
• Simply extend the previous models.
Cons
• Latest service would always be contained in a globals model folder to
prevent code duplication;
• Many more files; not easy to simply see what is new and what is not.
• Modules work for this but; only really work if your entire api is a
module and the other application is not. Which removes part of our
reuse.
40
42. Consumer Performance
• Planning for consumer performance is important for
instance:
Setting a reasonable timeout on the client end.
Messaging the user if it did timeout or some other issue occurred.
Utilizing some form of caching between the service and you
(something like a squid proxy) to help boost the requests per
second.
Implementing local caching and keeping track of things for that
user.
Utilize parallelism; if you need to grab multiple things do them at
one time.
Parsing is expensive; do it one and pass it around.
42
43. Who are you?
• When implementing web services; there are a few
additional things to think about
Rate Limiting
• Implemented more or less in a plug-in.
Authentication
• Leveraging tokens or API keys.
– A token service can deal with authentication and even allow them to have
sessions through the service layer.
Authorization
• Extending the server components to be aware of resources.
– This is harder but gives fine grained control.
• Otherwise, if you allow to a service they can access the entire service.
43
44. Zend Framework 2
• How will this all work in ZF2?
Well; I’m not fully sure but there is an RFC!
• http://framework.zend.com/wiki/display/ZFDEV2/RFC+-
+Server+Classes
All I can say; is it will be better with the new EventManager
• Authorization will become easier.
• Plugging in will be entirely more flexible.
44