Lean Enterprise with Microservices and Big Data

Transcript

1. How to enable the Lean Enterprise Johann Romefort co-founder @ rainbow
2. My Background • Seesmic - Co-founder & CTO Video conversation platform Social media clients…lots of pivots :) • Rainbow - Co-founder & CTO Enterprise App Store
3. Goal of this presentation • Understand what is the Lean Enterprise, how it relates to big data and the software architecture you build • Have a basic understanding of the technologies and tools involved
4. What is the Lean Enterprise? http://en.wikipedia.org/wiki/Lean_enterprise “Lean enterprise is a practice focused on value creation for the end customer with minimal waste and processes.”
5. Enabling the OODA Loop ! ! “Get inside your adversaries' OODA loop to disorient them” ! OBSERVE ORIENT DECIDE ACT USAF Colonel John Boyd on Combat: OODA Loop
6. Enabling the OODA Loop OODA Loop
7. The OODA Loop for software image credit: Adrian Cockcroft
8. OODA Loop • (Observe) Innovation and (Decide) Culture are mainly human-based • Orient (BigData) and Act (Cloud) can be automated
9. ORIENT
10. What is Big Data? • It’s data at the intersection of 3 V: • Velocity (Batch / Real time / Streaming) • Volume (Terabytes/Petabytes) • Variety (structure/semi-structured/unstructured)
11. Why is everybody talking about it? • Cost of generation of data has gone down • By 2015, 3B people will be online, pushing data volume created to 8 zettabytes • More data = More insights = Better decisions • Ease and cost of processing is falling thanks to cloud platforms
12. Data flow and constraints Generate Ingest / Store Process Visualize / Share The 3 V involve heterogeneity and make it hard to achieve those steps
13. What is AWS? • AWS is a cloud computing platform • On-demand delivery of IT resources • Pay-as-you-go pricing model
14. Cloud Computing + + Storage Compute Networking Adapts dynamically to ever changing needs to stick closely to user infrastructure and applications requirements
15. How does AWS helps with Big Data? • Remove constraints on the ingesting, storing, and processing layer and adapts closely to demands. • Provides a collection of integrated tools to adapt to the 3 V’s of Big Data
 • Unlimited capacity of storage and processing power fits well to changing data storage and analysis requirements.
16. Computing Solutions for Big Data on AWS Kinesis EC2 EMR Redshift
17. Computing Solutions for Big Data on AWS EC2 All-purpose computing instances. Dynamic Provisioning and resizing Let you scale your infrastructure at low cost Use Case: Well suited for running custom or proprietary application (ex: SAP Hana, Tableau…)
18. Computing Solutions for Big Data on AWS EMR ‘Hadoop in the cloud’ Adapt to complexity of the analysis and volume of data to process Use Case: Offline processing of very large volume of data, possibly unstructured (Variety variable)
19. Computing Solutions for Big Data on AWS Kinesis Stream Processing Real-time data Scale to adapt to the flow of inbound data Use Case: Complex Event Processing, click streams, sensors data, computation over window of time
20. Computing Solutions for Big Data on AWS RedShift Data Warehouse in the cloud Scales to Petabytes Supports SQL Querying Start small for just $0.25/h Use Case: BI Analysis, Use of ODBC/JDBC legacy software to analyze or visualize data
21. Storage Solution for Big Data on AWS DynamoDB RedShift S3 Glacier
22. Storage Solution for Big Data on AWS DynamoDB NoSQL Database Consistent Low latency access Column-base flexible data model Use Case: Offline processing of very large volume of data, possibly unstructured (Variety variable)
23. Storage Solution for Big Data on AWS S3 Use Case: Backups and Disaster recovery, Media storage, Storage for data analysis Versatile storage system Low-cost Fast retrieving of data
24. Storage Solution for Big Data on AWS Glacier Use Case: Storing raw logs of data. Storing media archives. Magnetic tape replacement Archive storage of cold data Extremely low-cost optimized for data infrequently accessed
25. What makes AWS different when it comes to big data?
26. Given the 3V’s a collection of tools is most of the time needed for your data processing and storage. Integrated Environment for Big Data AWS Big Data solutions comes integrated with each others already AWS Big Data solutions also integrate with the whole AWS ecosystem (Security, Identity Management, Logging, Backups, Management Console…)
27. Example of products interacting with each other.
28. Tightly integrated rich environment of tools On-demand scaling sticking to processing requirements + = Extremely cost-effective and easy to deploy solution for big data needs
29. • Error Detection: Real-time detection of hardware problems • Optimization and Energy management Use Case: Real-time IOT Analytics Gathering data in real time from sensors deployed in factory and send them for immediate processing
30. First Version of the infrastructure Aggregate Sensors data nodejs stream processor On customer site evaluate rules over time window in-house hadoop cluster mongodb feed algorithm write raw data for further processing backup
31. Version of the infrastructure ported to AWS Aggregate Sensors data On customer site evaluate rules over time window write raw data for archiving Kinesis RedShift for BI analysis Glacier
32. ACT
33. Cloud and Lean Enterprise
34. Let’s start with a personal example
36. First year @seesmic • Prototype becomes production • Monolithic architecture • No analytics/metrics • Little monitoring • Little automated testing
37. I built a monolith
38. or…at least I tried
39. Early days at Seesmic First year @seesmic
40. Everybody loves a good horror story
41. We crashed Techcrunch
43. What did we do?
44. Add a QA Manager
45. Add bearded SysAdmin
46. We added tons of process so nothing can’t go wrong
47. Impact on dev team • Frustration of slow release process • Lots of back and forth due to bugs and the necessity to test app all over each time • Chain of command too long • Feeling no power in the process • Low trust
48. Impact on product team • Frustration of not executing fast enough • Frustration of having to ask for everything (like metrics) • Feeling engineers always have the last word
49. Impact on Management
50. • Break down software into smaller autonomous units • Break down teams into smaller autonomous units • Automating and tooling, CI / CD • Plan for the worst What can you do?
51. = Break down software into smaller autonomous units
52. Introduction to Microservices
53. Monolith vs Microservices - 10000ft view -
54. Monolith vs Microservices - databases -
55. Monolith vs Microservices - servers -
56. Microservices - example -
57. Break down team into smaller units
58. Amazon’s “two-pizza teams” • 6 to 10 people; you can feed them with two pizzas. • It’s not about size, but about accountability and autonomy • Each team has its own fitness function
59. • Full devops model: good tooling needed • Still need to be designed for resiliency • Harder to test Friction points
60. Continuous Integration (CI) is the practice, in software engineering, of merging all developer working copies with a shared mainline several times a day
61. Continuous Deployment
62. Continuous Deployment
63. Tools for Continuous Integration • Jenkins (Open Source, Lot of plugins, hard to configure) • Travis CI (Look better, less plugins)
64. Tools for Continuous Deployment • GO.cd (Open-Source) • shippable.com (SaaS, Docker support) • Code Deploy (AWS) + Puppet, Chef, Ansible, Salt, Docker…
65. Impact on dev • Autonomy • Not afraid to try new things • More confident in codebase • Don’t have to linger around with old bugs until there’s a release
66. Impact on product team • Iterate faster on features • Can make, bake and break hypothesis faster • Product gets improved incrementally everyday
67. Impact on Management
68. • Enabling Microservices architecture • Enabling better testing • Enabling devops model • Come talk to the Docker team tomorrow!
69. Thank You follow me: @romefort romefort@gmail.com