What I Did On My Cloud Vacation

Summary after first six months of work on HUIT Cloud Engineering team

Transcript

1. What I Did On My Cloud Vacation March 3, 2014 • 60 Oxford St.
2. Exploring Strange New Worlds
3. • Tools/Outreach – Nepho (current release is 1.2.2) – Consulting (IAM, SIS, others) – Blog, lunch talks, office hours
4. • Tools/Outreach – Nepho (current release is 1.2.2) – Consulting (IAM, SIS, others) – Blog, lunch talks, office hours • Research IaaS Cloud – OpenStack on commodity hardware – Implementation by HUIT Cloud Engineering – Demo of Hadoop-aaS for James Cuff and Orran Krieger (BU)
5. • Tools/Outreach – Nepho (current release is 1.2.2) – Consulting (IAM, SIS, others) – Blog, lunch talks, office hours • Research IaaS Cloud – OpenStack on commodity hardware – Implementation by HUIT Cloud Engineering – Demo of Hadoop-aaS for James Cuff and Orran Krieger (BU) • Enterprise IaaS Cloud – RHEL OpenStack on Cisco UCS – Implementation by Red Hat – Pilot for Dev/Test use by enterprise customers
6. • Tools/Outreach – Nepho (current release is 1.2.2) – Consulting (IAM, SIS, others) – Blog, lunch talks, office hours • Research IaaS Cloud – OpenStack on commodity hardware – Implementation by HUIT Cloud Engineering – Demo of Hadoop-aaS for James Cuff and Orran Krieger (BU) • Web Application PaaS – Selected Red Hat OpenShift – Straightforward implementation path on OpenStack – Better solution for standalone web hosting • Enterprise IaaS Cloud – RHEL OpenStack on Cisco UCS – Implementation by Red Hat – Pilot for Dev/Test use by enterprise customers
7. Seeking Out New Life And New Civilizations
8. • Interdisciplinary teams of subject matter experts – Real-time collaboration – Rapid, incremental peer review – Toolchain sharing, “sharpening the saw”
9. • Interdisciplinary teams of subject matter experts – Real-time collaboration – Rapid, incremental peer review – Toolchain sharing, “sharpening the saw” • Automation is a tremendous force multiplier – 10% is better than nothing – 90% is not that much better than 10% – 100% is MUCH better
10. • Interdisciplinary teams of subject matter experts – Real-time collaboration – Rapid, incremental peer review – Toolchain sharing, “sharpening the saw” • Automation is a tremendous force multiplier – 10% is better than nothing – 90% is not that much better than 10% – 100% is MUCH better • Interruptions are terribly expensive – Engineering vs. operations work – Maker time vs. manager time – Who will you allocate?
11. • Interdisciplinary teams of subject matter experts – Real-time collaboration – Rapid, incremental peer review – Toolchain sharing, “sharpening the saw” • Automation is a tremendous force multiplier – 10% is better than nothing – 90% is not that much better than 10% – 100% is MUCH better • Build the system you want to work with – Deprecate legacy solutions – Don't let the edge case dictate policy – Fail fast • Interruptions are terribly expensive – Engineering vs. operations work – Maker time vs. manager time – Who will you allocate?
12. Boldly Going Where No HUIT Server Systems, Storage and Data Centers Team Has Gone Before
13. • Managing an IaaS platform is like managing a large, homogenous, integrated application – Enterprise support from vendor – Much higher degree of consistency/standardization/simplification – Automated management is not optional – Redundancy and fault tolerance is not optional
14. • Managing an IaaS platform is like managing a large, homogenous, integrated application – Enterprise support from vendor – Much higher degree of consistency/standardization/simplification – Automated management is not optional – Redundancy and fault tolerance is not optional • Managing instances on an IaaS platform is just like managing instances in the traditional datacenter – EXCEPT that we do everything programmatically – Think of instances as ephemeral – Marginal costs of additional instances/storage/network are minimal – Hardware specifications of instances are immutable – Load balancing for fault tolerance and flexibility is pervasive
15. • Allocate resources – We cannot do this work in 5 minutes here, 5 minutes there – 100% utilization means there is no time to learn new skills – Sending everyone to training doesn't accomplish anything if they all go back to the daily treadmill
16. • Allocate resources – We cannot do this work in 5 minutes here, 5 minutes there – 100% utilization means there is no time to learn new skills – Sending everyone to training doesn't accomplish anything if they all go back to the daily treadmill • Stop doing things by hand – Perform a searching and fearless inventory of your operational procedures – Become familiar with automation technologies – We are all developers now
17. • Allocate resources – We cannot do this work in 5 minutes here, 5 minutes there – 100% utilization means there is no time to learn new skills – Sending everyone to training doesn't accomplish anything if they all go back to the daily treadmill • Stop doing things by hand – Perform a searching and fearless inventory of your operational procedures – Become familiar with automation technologies – We are all developers now • Designate two team members in each team to become subject matter experts – Identify the AWS offerings that correspond to the work your team currently does – Send your SMEs to training, make time and space for them to learn – Hold them accountable for mastering the material and training the rest of the team
18. Q & A