Scene 1 (0s)
AMT Lesson Learned.
Scene 2 (5s)
What is AMT. Address Matching Tools Part of the four DITD Cloud Development Components for FY23 Address Matching Service Address Provisioning Service.
Scene 3 (15s)
How we got started? Page 1. Training Took lots of trainings from amazon. Very confusing. Too many new names and acronyms. Lots of hands-on Labs in the training Things start falling into places after a while.
Scene 4 (29s)
The Team. APSB/AUSB – The developer SME – The product owner. The requirement provider. QTMB – The software tester EWG - Provide the approval for moving forward with your cloud project DevSecOps – create CI/CD Pipeline MSS – Proxy to SCT and other cloud service provider DSB - Because of the serverless choices made, DSBs role was minimal. OIS - Integrate them into the process as soon as possible SCT Provided access to the cloud accounts and services interfacing Management - Much needed support with getting all groups to done..
Scene 5 (52s)
How we got started? Page 2. Research and Evaluation Lift and Shift vs Cloud Native decision EWG really shined in this area (i.e.: Athena and other serverless tech lambda, step, etc.).
Scene 6 (1m 5s)
How we got started? Page 3. Architect Ongoing Keeps Changing from learning Keeps Changing from collaborating - EWG Cost.
Scene 7 (1m 14s)
Lesson 1 – Just do it.. Start now. Keep designing the real application. It won’t be your only version. It will allow you to…. Start smaller to.
Scene 8 (1m 26s)
Our Solution. Cloud Native Cost-conscious – Athena, Lambda, S3, Step Functions, ECS What we learned Cheaper Performance hits for service startup How you use serverless makes difference.
Scene 9 (1m 37s)
Lesson 2 – Cost Saving. How you use serverless makes difference Change the way you think how function is called/used Traditional Function call – Function A calls B, B calls C You are paying for time used by Function A, B, and C AT THE SAME TIME Make it cheaper by using Step Functions Step Functions calls Function A Output from A is passed to B, and Function A is shut down. Output from B is passed to C, and Function B is shut down. Therefore, you pay only for time processing for each one Function at a time.
Scene 10 (2m 2s)
Lesson 3 – Performance. Use Athena, a serverless database, for our address matching data store. Implemented our Athena Matching Function just like on-prem known as Oracle Stored Procedure. Bad performance – 9 hours for a testcase Re-Implemented Athena Matching Function as an independent Lambda Function. Performance improved from 9 hours to 20 mins. Lesson Learned Athena Matching Function is called sequentially and didn’t run in parallel Lambda function allows Multi-Threading.
Scene 11 (2m 22s)
Key Points. How you develop on-prem does not always translate to how it should be done on cloud..
Scene 12 (2m 32s)
Security Security Security. As a developer, Security Concern is NOW your concern By default, you have permission to do NOTHING. Lambda Function needs permission to talk to Athena and S3 buckets API Gateway needs permission to execute Lambda Athena needs permission to use Glue All services need permission.
Scene 13 (2m 47s)
Security Security Security. As part of the release process for code to go from DEV to TEST and PROD CI/CD Pipeline Approval Gates PENTEST Plan for extra time to release your software, especially if this is the first time Documentation Architecture User Management Vulnerability Report ……. Title 13 No MAFTIGER@devtran in the cloud.
Scene 14 (3m 4s)
Agile is your friend. Able to adjust in any schedule delay Allow flexibility in continuous improvement on requirement/user stories Allow adding more people to support the project and catch up on the schedule Allow participation from different areas with continuous collaboration – security, middleware, testers, product owner, developers ….