Turning pets into cattle - the stickiness of data Dr Yih Leong Sun, - - PowerPoint PPT Presentation

▶
turning pets into cattle
SMART_READER_LITE
LIVE PREVIEW

Turning pets into cattle - the stickiness of data Dr Yih Leong Sun, - - PowerPoint PPT Presentation

May 22, 2023 231 likes •505 views

Turning pets into cattle - the stickiness of data Dr Yih Leong Sun, Chris Cannon, Gerd Prmann April 26th, 2016, 3:40pm-4:20pm Introduction Dr Yih Leong Sun Chris Cannon Gerd Prmann Senior Cloud Architect Cloud Engineering Manager

slide-1
SLIDE 1

Turning pets into cattle

  • the stickiness of data

Dr Yih Leong Sun, Chris Cannon, Gerd Prüßmann April 26th, 2016, 3:40pm-4:20pm

slide-2
SLIDE 2

Introduction

Dr Yih Leong Sun Senior Cloud Architect Intel Gerd Prüßmann Director Cloud Solutions Mirantis Chris Cannon Cloud Engineering Manager Hewlett Packard Enterprise

Reference: https://wiki.openstack.org/wiki/Enterprise_Working_Group

slide-3
SLIDE 3

What is this session NOT about?

Continuous Integration and Continuous Deployment Automation Techniques for Infrastructure, Deployment, and Operations Agile Development Practices It’s an application architecture discussion

slide-4
SLIDE 4

Agenda

  • Recap of previous presentation
  • The stickiness of data

– Data challenges – Strategies

  • Q & A
slide-5
SLIDE 5

Previous presentation

slide-6
SLIDE 6

Pets vs Cattle

Reference: https://www.openstack.org/assets/Uploads/presentation-turning-pets-into-cattle.pdf https://www.openstack.org/summit/tokyo-2015/videos/presentation/turning-pets-into-cattle-a-demonstration-to-provoke-discussion

slide-7
SLIDE 7

Reference: http://www.openstack.org/assets/pdf-downloads/business-perspectives.pdf

Virtualization vs Cloud

slide-8
SLIDE 8

Architecture design

Conventional App

  • Monolithic
  • Centralised state
  • Tightly coupled
  • Synchronous
  • Single tenancy

Cloud-aware App

  • Distributed
  • Microservices
  • Asynchronous
  • Multi-tenancy
  • Failure-resilience
  • Share-nothing
  • Eventually

consistent

vs

Reference: http://www.opendatacenteralliance.org/docs/architecting_cloud_aware_applications.pdf

slide-9
SLIDE 9

Multiple strategies

VM VM VM Web App DB Web App VM VM VM Web App DB w1 w2 a1 a2 w1 a1 static content

slide-10
SLIDE 10

Data stickiness

slide-11
SLIDE 11

Data challenges

  • Every Pet is different
  • Challenges of existing data

– How to migrate existing data to the cloud?

  • Data migration: file copying, hot/cold export, machine copying,

continuous replication …

  • Integrity, consistency, licensing & down time

– Data virtualization

  • SLA (policies, performance, RTO, RPO, etc ...)
  • What is your current situation and limitation?
slide-12
SLIDE 12

Data strategies

slide-13
SLIDE 13

Option 1 - No change to backend data store

  • Lower migration risk by using existing DB deployments
  • Most PaaS systems can connect to legacy DB

environments

  • Continue to leverage capabilities that may not be fully

available in cloud-based DB today

slide-14
SLIDE 14

Option 2 - ‘Cloudifying’ RDBMS / Example

*Other names and brands may be claimed as the property of others.

Pros

  • Known technology / operational tasks
  • Leverages DB scalability

Cons

  • Limitations / reqs to the schema
  • Not all engines available
  • Operational tasks remain
  • Automation efforts
slide-15
SLIDE 15

Option 2 - ‘Cloudifying’ RDBMS / Trove

  • Openstack DBaaS project
  • ffering self-service provisioning

functionality for RDBMS and non-RDBMS

  • easy on-demand DB

consumption

  • multiple DB instances per user
  • resource isolation
  • task automation for

deployment, config, patching etc.

Cassandra, CouchBase, CouchDB, DataStax Enterprise, DB2, MariaDB, MongoDB, MySQL, Oracle, Percona Server, PostgreSQL, Redis and Vertica

slide-16
SLIDE 16

Option 2 - ‘Cloudifying’ RDBMS / Trove

Key Features Capability Client Database Performance General Purpose (SSD) Storage Yes - Relies on OpenStack Cinder storage capability Provisioned IOPS (SSD) Storage Yes - Relies on OpenStack Cinder storage capability Scalability Push-button Compute Scaling Limited - Requires manual expansion Easy Storage Scaling Limited - Requires migration to new database Availability and Durability Automated Backups Limited- Requires external backup solution Database Snapshots Limited - Requires external backup solution Multi-AZ Deployments Yes - Limited to a single OpenStack environment. No option to specify specific AZ. DR with Multi-Site/multi-Region Deployment No Automatic Host Replacement No Security Encryption at Rest and In Transit Limited Network Isolation Yes Resource-level Permissions Yes Manageability Monitoring and Metrics No - Requires external monitoring solution. No SLA calculation. Automatic Software Patching No - Requires external patching solution DB Event Notifications No - Requires external notification solution Metering Collect usage metrics for billing No current usage metrics are emitted by Trove

slide-17
SLIDE 17

Example: WordPress*

  • Use alternative ‘cloud-native’ DB backends for WordPress ?

– Official WordPress supports MySQL* / MariaDB* engines only – PostgreSQL* ongoing – NoSQL backend not planned

  • Codebase: MySQL centric
  • Approaches:

– Status Quo: Development for MySQL only

  • Fork necessary for every new database type (costly for porting)

– WordPress specific database abstraction layer – Full database abstraction layer

Reference: https://codex.wordpress.org/Using_Alternative_Databases

*Other names and brands may be claimed as the property of others.

slide-18
SLIDE 18

Option 3 - RDBMS to NoSQL

  • Why should I migrate from RDBMS to NoSQL?

Database Example Use Case NoSQL Cassandra*, MongoDB* IoT, BigData, Widely distributed architecture unstructured data RDBMS MySQL*, MSSQL* OLAP, OLTP, ACID, structured data

Reference: https://www.mongodb.com/collateral/rdbms-mongodb-migration-guide

*Other names and brands may be claimed as the property of others.

slide-19
SLIDE 19

Option 3 - RDBMS to NoSQL

  • Migration tactics

– Build your own scripts, ETL, Hadoop – Snapshot and incremental – Application-driven

  • Keeping data in both RDBMS & NoSQL

concurrently during transition period

  • On-demand
slide-20
SLIDE 20

Summary

  • Understanding your Pets
  • Some existing database technologies are not designed for Cloud
  • Differences in design principle between RDBMS & NoSQL
  • Transformation from existing environment to Cloud ‘culture’
  • Skillsets and training for Ops and Dev
slide-21
SLIDE 21

Q & A

slide-22
SLIDE 22

OpenStack Enterprise Working Group

  • Objective:

To identify and remove barriers to Enterprise adoption and deployment of OpenStack*.

  • Mailing List:

http://lists.openstack.org/cgi-bin/mailman/listinfo/enterprise-wg

  • Meeting Logistics:

Thursday 1700 UTC Call In Info: (888) 875-9370, Bridge: 3; Passcode: 9518007#

JOIN US!

*Other names and brands may be claimed as the property of others.

slide-23
SLIDE 23

Thank you

yih.leong.sun@intel.com ccannon@hpe.com gpruessmann@mirantis.com

slide-24
SLIDE 24

Legal Disclaimer

OpenStack is a registered trademark of the OpenStack Foundation in the United States, other countries or both. Intel and the Intel logo are trademarks of Intel Corporation in the U.S. and/or

  • ther countries.

Mirantis, and the Mirantis Logo are registered trademarks that belong to Mirantis Inc.

slide-25
SLIDE 25