Cloud-native persistence in a serverless world
0 likes278 views
The document discusses cloud-native applications and their persistent characteristics, focusing on serverless technology and various storage options such as relational and non-relational databases, object storage, and message queues. It provides case studies, particularly on the education platform edX, to illustrate different persistence options and considerations for cloud services. Additionally, it covers best practices for operational expertise, compliance, and the impact of choosing managed versus self-hosted services.
Cloud-native persistence in a serverless world
- 2. > print(“Hello World!”) ✓ 1+ Year of experience using AWS ✓ 2 x AWS certified ✓ Also love to chat with your about electromagnetics, magnets & motors ✓ Love learning & sharing Serverless Meetup Auckland – 17/04/18
- 3. What I want to cover ✓ Cloud-native applications ✓ Persistent characteristics ✓ Choice of approach - on the Serverless scale ✓ Cloud-native persistent options – OpenEdx example ✓ Considerations Serverless Meetup Auckland – 17/04/18
- 4. 1. Cloud-native applications Utilize benefits of the cloud ScaleDurability Availability Serverless Meetup Auckland – 17/04/18
- 5. 1. Cloud-native applications DecoupledSmall Stateless Tends to follow Microservices architecture Serverless Meetup Auckland – 17/04/18
- 6. 2. Persistent characteristics ✓ Meta data ✓ Documents (XML, YAML, JSON) ✓ Stream (Logs, Time series…) ✓ Object (Media, sound..) ✓ Transactional queries Balancer Web Server Reverse Proxy Application Server Data type = format for encoding data Serverless Meetup Auckland – 17/04/18
- 7. 2. Persistent characteristics Consistency model = rules of memory operation for each UPDATE https://www.slideshare.net/springerw/eventually-consistent Serverless Meetup Auckland – 17/04/18
- 8. 2. Persistent characteristics Persistency = the keeping of data that you are working with https://briantroy.com/2010/11/02/big-data-storage-isnt-enough/Serverless Meetup Auckland – 17/04/18
- 9. 2. Persistent characteristics Latency = time taken to perform a READ or WRITE ms seconds minutes hours Caches NoSQL SQL HDFS Object Depends on size Serverless Meetup Auckland – 17/04/18
- 10. 2. Persistent characteristics ✓ Data type ✓ Consistency model ✓ Persistency ✓ Latency Concurrency Streaming capability Availability/SLA Durability … Cassandra Serverless Meetup Auckland – 17/04/18
- 11. 3. Choice of approach Managed services Serverless services only accessible via APISelf-host (DIY) Scaling delay • You control • Within minutes or hours • Instant or within seconds Pricing model • Servers up time • Cold storage & bandwidth • Pay per request Security & Compliance • Implemented with best practices • Rely on Compliances offered by the provider • You control • Invulnerable to classic attacks (cross site scripting, injections…) • Rely on Compliances offered by the provider Control • Full control (reaction speed, optimization, tenancy, compliance) • Highly restrictive • Leave you with control of scaling, encryption & backup • No control, not even scaling Best fit • High resource utilization • Predictable demand • Strict SLA & Compliances • Operation expertise • Focus on growth • Can afford some operation work • Focus on growth • Loose latency requirements • Unpredictable demand • Peace of mind! Serverless Meetup Auckland – 17/04/18
- 12. 4. Persistence options – case study: • edX: non profit organization founded by MIT & Havard • Open edX: LMS & MOOC platform released in June 2013 • Theoretically possible: 1,000,000 concurrent users • Powers edx.org & cousera.com Serverless Meetup Auckland – 17/04/18
- 13. 4. Persistence options – case study: edx.org & cousera.com Serverless Meetup Auckland – 17/04/18
- 14. VM local & Block storage Amazon EC2 Amazon EBS & GCP Compute Engine GCP Persistent Disk & Workload Management • Very low latency disk • High IPOS • High throughput • Self-managed Example • Web, application server • Literally anything Azure Virtual Machine Azure Disk Storage & 4. Persistence options – case study: Serverless Meetup Auckland – 17/04/18
- 15. Object Storage Workload Management • Static object storage • Static web hosting • Static content distribution • Serverless Example • Static website • Video, audio, document GCP Cloud Storage Amazon S3 Azure Storage - Blobs 4. Persistence options – case study: Serverless Meetup Auckland – 17/04/18
- 16. File Storage Workload Management • Shared file system • Low latency access • Access semantic required (strong consistency, locking) • Managed Services Example • CMS Template storage • Configuration management Amazon EFS Azure Storage - Files GCP Persistent Disk (multiple VM, read-only) 4. Persistence options – case study: Serverless Meetup Auckland – 17/04/18
- 17. Message Queue Workload Management • Pull-based delivery • Decoupling components • Serverless Example • Processing request queue GCP Task Queue Amazon SQS Azure Storage - Queue 4. Persistence options – case study: Serverless Meetup Auckland – 17/04/18
- 18. Relational database Workload Management • OLTP/OLAP • Structured data • SQL Query capability • Managed Services Example • Student details • Sessions ID • Data warehouse & analytics Amazon RDS & Aurora Amazon Redshift Azure SQL Database Azure SQL Data warehouse GCP Cloud SQL GCP Cloud Spanner GCP BigQuery 4. Persistence options – case study: Serverless Meetup Auckland – 17/04/18
- 19. Non-Relational database Workload Management • OLTP • Flexible structured data • Highly scalable • Managed Services Example • Course contents • Forums GCP Cloud Datastore GCP Cloud Bigtable Amazon DynamoDB Azure CosmoDB 4. Persistence options – case study: Serverless Meetup Auckland – 17/04/18
- 20. In-memory caches Workload Management • In memory key-value store • Database Query result cache • Compute result cache • Managed Services Example • Session information • Save READ heavy workload CDN Workload Management • Static content distribution • Access controlled static content distribution • Managed Services Example • Reduce static website load speed GCP Cloud CDN Amazon CloudFront Azure CDNAzure Cache Amazon Elasticache Amazon DynamoDB Accelerator 4. Persistence options – case study: Serverless Meetup Auckland – 17/04/18
- 21. 4. Persistence options – NEW: AWS Neptune Graph database Workload Management • OLTP/OLAP • Highly connected data • Managed Services Example • Knowledge graph • Social networking • Recommendation engine https://www.awsgeek.com/posts/amazon-neptune-notes/ https://aws.amazon.com/neptune/Serverless Meetup Auckland – 17/04/18
- 22. 4. Persistence options – NEW: Serverless Aurora • Serverless SQL database • Upto 64TB • SQL Workload: • Intermittent • Unpredictable • Rapid scaling • Pay for cold storage & requests https://aws.amazon.com/blogs/aws/in-the-works-amazon-aurora-serverless/ Serverless Meetup Auckland – 17/04/18
- 23. 5. Considerations - technical • Data Format • Consistency model • Persistency • Latency • Concurrency • Streaming capability • Availability/SLA • Durability … JSON object • Read • Write Use the simplest type of storage … ALLOWED Simple Read & Write Object Storage File Storage Consistent latency OLTP Databases Data warehouse Serverless Meetup Auckland – 17/04/18
- 24. 5. Considerations - business ✓ Self-host (DIY) ✓ Managed services ✓ Serverless Compliances requires LOTS of work … compliant cloud applications requires compliant components. AWS Lambda got PCI DSS compliance (Payment Card Industry) in June 2017 AWS RDS got HIPAA compliance (Healthcare data) in Sep 2017 https://aws.amazon.com/blogs/security/aws-adds-12-more-services-to-its- pci-dss-compliance-program/ Serverless Meetup Auckland – 17/04/18 https://aws.amazon.com/about-aws/whats-new/2017/09/amazon-rds-for-sql- server-and-amazon-rds-for-mariadb-achieve-hipaa-eligibility/
- 25. 5. Considerations - business ✓ DevOps ✓ Managed services ✓ Serverless Starts with NO operation … initially DropBox started with just UI on top of Amazon S3. https://www.slideshare.net/ChiaraCilardo/dropbox-architecture-and-business-prospective Serverless Meetup Auckland – 17/04/18
- 26. 5. Considerations - business ✓ DevOps ✓ Managed services ✓ Serverless DevOps is NOT going away soon … instead of fully embracing 1 approach: analyze, try , and gain. https://medium.com/openwhisk/uncovering-the-magic-how-serverless-platforms-really-work-3cb127b05f71 https://github.com/openfaas/faas IBM OpenWhisk – Serverless Computing is built on top of DevOps OpenFaaS – Serverless Functions with first class metric support Serverless Meetup Auckland – 17/04/18
- 27. Thank you for your time! Any questions?
- 28. Bonus: Key-value vs. Object vs. Block vs. File https://blog.rackspace.com/introduction-to-object-storage Key-value vs. Object storage: • Both access using API calls • Object storage allow you to look into and query the content Block/Blobs storage: • SAN, iSCSI, RAID … • The lowest level + attach directly to OS File storage: • CIFS or NFS • Still a file system like Block/Blobs, but indirect access via translated command Serverless Meetup Auckland – 17/04/18
- 29. Bonus: Should I use Amazon DynamoDB? https://read.acloud.guru/why-amazon-dynamodb-isnt-for-everyone-and-how-to-decide-when-it-s-for-you-aefc52ea9476 Serverless Meetup Auckland – 17/04/18
- 30. Bonus: GCP data store – a selection guide https://medium.com/google-cloud/a-gcp-flowchart-a-day-2d57cc109401 Serverless Meetup Auckland – 17/04/18
- 31. I wish to acknowledge: • Mr. Andreas Mueller – General Manager at API Talent • Mr. Kevin Tran – Senior Software Architect, Tech Evangelist • Mr. Nghia Nguyen – Senior Solution Architect at Datacom Honorable mention: • Mr. Harel Aharony - eCommerce Platform Lead at XM Developments References: • https://medium.com/google-cloud/a-gcp-flowchart-a-day-2d57cc109401 • https://medium.com/openwhisk/uncovering-the-magic-how-serverless-platforms-really-work-3cb127b05f71 • https://medium.com/@james.anderson_88068/aws-first-look-advancements-in-aurora-fcb7ff4c8b9e • https://read.acloud.guru/the-serverless-spectrum-147b02cb2292 • https://read.acloud.guru/simon-wardley-is-a-big-fan-of-containers-despite-what-you-might-think-18c9f5352147 • https://thenewstack.io/continuum-cloud-native-topologies/ • https://aws.amazon.com/blogs/aws/in-the-works-amazon-aurora-serverless/ • https://read.acloud.guru/why-amazon-dynamodb-isnt-for-everyone-and-how-to-decide-when-it-s-for-you-aefc52ea9476 Good reads I wish to share: • “The Phoenix Project” by Gene Kim, George Spafford, and Kevin Behr • “Serverless Architectures on AWS” by Peter Sbarski • “Architecting for Scale” by Lee Atchison /in/nick-do/ nick.do@datumhq.com Serverless Meetup Auckland – 17/04/18