Tuesday, May 24, 2016
Passing the Salesforce Certified Integration Architecture Exam
I recently passed the Salesforce Certified Integration Architecture Exam. The full exam details can be found at the SFDC Architect Academy. This is one of the 9 Architect Domain Specialist exams. In this post I'll be providing my general thoughts on the content of the exam. Here is the description of the exam from the Architect Academy:
The exam asks multiple scenario based questions where you need to choose the right integration pattern given a set of criteria. For example a question may look something like:
Delivery Express has a Salesforce.com Org that is holds all the customer information including Accounts and Opportunities. Delivery Express has a on-premise ERP system that is the order fulfillment system. The ERP System has a SOAP API for creating and updating orders and is offline occasionally for maintenance. Delivery Express has a new requirement whenever an Opportunity is Close-Won an order should be created in the ERP. The order needs to be near real time and custom development work is not an option.
Which technology should the Integration Architect select based on the above?
A. Outbound Messaging
B. Apex Callout
C. ETL Tool
D. Canvas App
The above question can be answered with A because Outbound Messaging can be setup custom development, can retry failed messages automatically, and supports SOAP API endpoints.
You need to know when to use a mashup. If the requirement is to simply view the data in SFDC UI then you can look at doing Canvas or Custom Web Tab. Use a Canvas if you need authentication and Custom Web Tab if you need to surface an external website that is unauthenticated.
You need to know when to use ETL and Middleware to solve your problem. Do you need complex error handling and orchestration across multiple systems? Then you're going to need to use an ESB, Middleware, or ETL tool. Do you have a data synchronization that does not need to be real time but handle large data volumes? Again an ETL tool using Bulk API is going to be your best bet.
The answer is not to always do the work on the SFDC Tier. Many times you want to pre-process your data before you load it into SFDC. One example is to group your data by a parent record like Account so that Bulk API jobs do not fail due to parent record locking. Also for example if you are integration large volumes of data you do not want to do upsert you want to do separate insert and update calls respectively, forcing you to pre-process those ID's in your data set.
I am currently looking for the next set of Force.com Engineers to mentor. If this type of technical content and challenge seem appealing to you, and you want to become an all star Force.com Engineer across the entire Force.com PaaS Stack, reach out to me on Twitter or Linkedin. I am looking for several USA based engineering roles to join my engineering team at Fusion Risk Management (Sorry I cannot provide visa sponsorships ).