2016 - Open Mic - IGNITE - Open Infrastructure = ANY Infrastructure
- 1. Open Infrastructure = ANY Infrastructure Hybrid DevOps on and Beyond OpenStack Rob Hirschfeld, CEO RackN Cloud + Data Center Automation for Hybrid Infrastructure
- 2. Open Ops: User Choice & App Portability Goal: run a reference workload (Kubernetes) on any infrastructure using the same operational process. Execution: a single command to run Kubernetes on OpenStack, Amazon, Google and Metal (via Packet. net) with SDN & O/S choices.
- 3. Demo #1 Multi-OpenStack To make things portable, we need to be able a repeatable experience between multiple clouds. We need to get similar resources from multiple OpenStack Clouds to run the same workload on multiple clouds.
- 4. Demo #2 Multi-Kubernetes To make things portable, we need to be able a repeatable experience between multiple clouds. We want to be able to run the same workload on multiple clouds from different vendors.
- 5. Hybrid is an overloaded term! Multiple Hybrid Dimensions: ● Different Vendors ● Different Platforms ● Different APIs ● Different DevOps Tools ● Different Operating Systems We’re talking about using infrastructure in change tolerant way. The only predictable thing about infrastructure is that is will change. Hybrid acknowledges that you will be using old and new and new new.
- 6. Hybrid Infrastructure is the new normal TWO THIRDS OF ENTERPRISES HAVE HYBRID IT (IDC)
- 7. And Infrastructure choice is increasing and many others... and many others... AWS GCE Azure RackSpac e and many others... “Bare Metal” On average, large enterprises are using about two dozen cloud services from nine providers (Gartner) VMware OpenStack OpenStack Public Private
- 8. Tools do not manage Hybrid IT - not just cloud, but ALL Infrastructure Cross-Platform Orchestration (aka Hybrid DevOps) fills gaps left by current ops tools and many others... and many others... AWS GCE Azure RackSpac e and many others... “Bare Metal” VMware OpenStack OpenStack “Why is it so hard to scale up this infrastructure?” “We need clawback our apps from AWS” “Data locality means I need data centers all over the world” “I need to consolidate data centers. How do I simplify management too?”
- 9. So…. we need a single API!
- 10. No! No single API or Platform wins.
- 11. Infrastructures have unique requirements Platforms WorkloadCloud, Physical & NetworkPhysical Infrastructures Step 2 Step 6 Step 7 Step 1 Step 3 Step 4 Step 2 Step 7 Step 1 Step 3 Step 4 Step 8 Step 10 Step 2 Step 1 Step 3 Step 5 Step 9 Step 11 Step 6 Step 9 Step 11 Step 6 Step 7 Step 4 Step 8 Step 9 Step 10 Step 11 Step 2 Step 6 Step 7 Step 1 Step 3 Step 4 Step 5 Step 8 Step 9 Step 10 Step 11 ApplicationWorkloads “Bare Metal” Ops need to create a system-wide control fabric by composing lots of individual actions in sequence Orchestration Step 5 Step 5 Step 8 Step 10 Step 2 Step 3 Step 4 Step 1 Step 2 Step 1 Step 3 Step 4 Step 2 Step 1 Step 3 Step 6 Step 7 Step 7 Step 8 Step 10 Step 5 Func Role Func Role Func Role Func Role Func Role Func Role Func Role Func Role Func Role Func Role Func Role
- 12. To Vendors: AWS Drives Operational Patterns AWS Azure GCE IBM RAX DO Amazon is so dominant in infrastructure that their patterns (API and Implementation) must be factored into any operational discussion. Even if it is a physical only deployment. Our hybrid DevOps objective is simple: We need multi-infrastructure Amazon equivalence for ops automation. This trend will accelerate an AWS competitor work to reduce switching friction off AWS. It is easier to recruit cloud users from AWS than IT Ops.
- 13. To Enterprise IT: AWS is disruptive but not only choice While AWS dominates the market, individual companies have a much more mixed infrastructure. They are starting from existing workloads. There are many factors for IT in infrastructure vendor choice including relationships, control and cost. When, mono-infrastructure is dead then portability becomes critical. AWS still sets the operations standard and that ultimately influences back into internal IT. AWS Alternate Public Vendor Cloud Private Cloud Internal IT
- 14. What makes Hybrid hard? Beneficial Diversity.
- 15. It may not be pretty, but working Ops is not wrong There are many ways to run infrastructure. Just because it’s different (or last generation) does mean that it’s wrong. Burning down your data center is not an effective option. Most operators would happily migrate to new tools if it was less disruptive. The alternative is to create more operational silos.
- 16. OpenStack lessons? API is important, but Implementation matters
- 17. It’s not just API, It’s Implementation No cloud tested would work with just the information from the openstack.rc file provided. Most clouds required OpenStack Floating IPs for external connectivity - this model breaks AWS & Google compatibility. External IP, Configuration and O/S are the most critical. Other items can be easily handled by site-specific configuration. All these items ARE surmountable but they are all distractions for users. External IP Complete Config Standard O/S Network Names Standard Login ✔ ✘ ✘ ✘ ✔ ✘ ✘✔ ✔ ✔ ✔✘ ✔ ✔ ✘✔ ✔ ✔ ✘✔
- 18. Operations drives Infrastructure Software Hardware Ops When I worked for Dell, we thought we could sell Scale Cloud and Big Data by just bundling them with some servers. Scale platforms have very high operational requirements and require automation. This is especially true because the platforms have sub-six month release cycles. Selling hardware or software without and operational story will frustrate customers.
- 19. Hybrid DevOps This is not just technology! Good hybrid design is about process, discipline and culture. We cannot rely on Configuration Mgmt to create portability. The current patterns create brittle towers of vertically wired automation. Robust designs require a composable modular design. Composable designs require orchestration for action chaining. Gets Most Focus Biggest Gap
- 20. Data Center Ops APP Hybrid Needs Composable Parts Deployments are always composed of a lot of moving parts. They are both integrated both vertically and horizontally (not shown). So incremental changes will dstrupt the whole stack. Everything is always changing. Robust deployments must be build with composable modules so that they can be fault tolerant and resilient to change. It is very expensive to add composition afterwards! Mgmt Tools Logical Net Operating System Infrastructure Provisioning APP Mgmt Tools Logical Net Operating System Infra- structure Provision- ing DC Ops Fragile Mono- Integration Interchangeable Composition Mgmt Tools Logical Net Operating System Infra- Structure as a Service
- 21. In Summary: ● Hybrid Infrastructure is new normal ● Operations can work Hybrid ● Amazon is the Ops benchmark ● Implementation choices matter ● Invest in making DevOps composable My blog http://RobHirschfeld.com @zehicle on Twitter