Tour de Clouds: Understanding Multi-Cloud Integration

Editor's Notes

• #3: IntroYears @ RSGroups3y ago in the new yorker…manifestogateIn the beginning…
• #5: In the beginning…
• #7: In the beginning…
• #10: RS Cloud Integration DeliversBasic infrastructure on which all RightScale functionality appliesMultiCloud Marketplace with ServerTemplatesGovernance / ComplianceAutoscaling, Monitoring, Alerting
• #11: Inundated with oversimplified representations of technology in the airport.Everything and everyone is in the cloud.EVERYONE says they have high availability, fault tolerant systems.EVERYONE has a deep roadmap of real or planned cloud service offerings aimed broadly: services that feel like appliances (ELB, RDS), use cases from big data to PCI compliance, audiences from developers to sys admins to managers.EVERYONE has an API. APIs are a given…a must. “yes we can provide that info…we keep that in our database”…when I hear that I think you might as well send that to us in a csv file that I can open in excel…that’s about how usable it is.HP conversation; Reliable IaaS has become a given….on-demand, reliability, consistency, automation, and most importantly, works…ALL of the time.…But theres more to it. It’s the behaviors and …its what the provider does with it that’s the differentiator.One viable cloud: this is not the highlander…there can be more than one for you.
• #13: Common images available that are vanilla and ready to automate on top of; common interface with standard and consistent cloud terminology through RS, we provide a single API that abstracts the nuances from each cloud provider’s API (translations for single API calls and possibly multiple API calls for some clouds); We take care of translating to the cloud specific API to perform the actions.Sometimes that’s 1:1 and sometimes that’s 1:many actions we abstract away. Ex, A volume is a volume is a volume no matter which cloud.Ex, cloud networks are the clouds network no matter which cloud.
• #14: How did we choose our partners: hard work…a lot of progress in 6 years…blood sweat and tears, high value and low-value investment in a constantly moving landscape. Would love to say it was 100% strategic, but it wasn’t wholly. Healthy mix of strong leadership, real-world experience, product vision, amazing partners, well-timed market opportunism. Right place. Right time.Its not academic. Against the backdrop of a volatile customer base and use case set…web 2 to enterprise. Web apps, social apps, batch processing, BI and big data, enterprise CRM, and the dreaded legacy application footprint in the basement. We focus on real-world production operational and management use cases, not one-time operations and POCs.We have a triple-digit list of cloud providers in various stages of enforceRightScale compliance/support…we’ve learned a lot along the way and are learning more every day.In the last 3 years, we’ve assembled a cloud provider pipeline in the triple-digits providers…pure play hardware companies, software providers with excess infrastructure, small and large hosting providers…and even book companies.Behind our product, we are a group of cloud computing consultants armed with a scalable process to evaluate and quickly size up public clouds…heres a look into our process:
• #15: Biggest enemy to cloud migration?Onboarding…55K users, triple digit onboarding engagements: Here’s a little-known fact. The vast majority of delays are unrelated to cloud or RS functionality. It’s the app.Common scalability blockers: Stateful appsDB-intensive, no use of cachingReliance on custom OS, custom kernelsDevelopers are not necessarily sys admins. Sys admins are not necessarily developers. Experienced Devops resources are becoming more plentiful, but hiring talent is the most difficult thing about what anyone in this room does. Solution? Invest in education. Invest in partners that will educate you as part of their product delivery. Standardize your IaaS. (Chef, Solution Set story). Live on the edge…a little.Most importantly: know what you do. You’re a 5-man dev shop with a killer app running on a server in your garage? Awesome, but maybe managing IaaS isn’t your gig. And that’s ok…almost all of the cloud providers offer varying offerings of human services in addition to cloud services. Managed Cloud Hosting. Managed Private Cloud Hosting. Managed Database Hosting. There are many options available that let you get all of the advantages of cloud…without.
• #16: We start with getting an understanding of a providers foundation. Lets start with hardware. It may come as a surprise that we don’t sweat it. Anyone can buy hardware…its not a competitive differentiator. What you do with this hardware is whats important.One exception to this…IOPSContrary to cloud’s promised use of commoditized hardware, extensive use of ONLY commodity hardware is atypical limited primarily to a cloud’s compute layer. Though there are a few well-known exceptions to this, increasingly we are seeing cloud partners invest primarily in high-speed network fabric, network interfaces, and complex storage solutions that leverage SAN, ISCSI, and SATA (duke’s email).Virtualization layer: we see primarily 4…Xen/Server, ESXi, KVM, Hyper-V. There are some considerations here…mainly for us. We’ve seen some rough edges around volumes, networking, and user constructs. IP addressing…as an example. Each virtualization technology presents them differently…we abstract them so ou don’t have to. Why is this important? We removed a criteria in your decision-making.APIs…machine resources – vms, images, instance types, volumes, storage, security groups, etc. only part of the story. Everyone exposes these. The devil is in the details…its about the behavior.
• #17: Regions and Azs, GeographyWhy is this important?As a cloud consumer, I want to architect my application for failure…some clouds make it easier than othersConsiderations: 3 words: failure isolation boundariesZones / DatacentersRegions / Geographiesresource availability across isolation boundariesImages to launch new instances should be global availablePortable volumes should be available across zones / DatacentersVolume snapshots should be globally availableNetworking: Public to private space – private to privateregion-to-regionConnection to the mothership: do you need VPN? does it need to be an integral service (hardware support) vs software-based…log data,Hybrid Cloud (vlan vs. flat)Network isolationCloud-to-dedicatedCloud services: vertically integrated, IaaS-enabled offerings…ELB,PaaS, DNS, CDN, etcKnow your users. Have a lot of developers w/o sys eng experience…consider providers with a strong set of PaaS or PaaS-like offerings. Have a deep devops bench…save on self-service. Concerned about lock-in? think hard about cloud-specific services…diversify?Human servicesIncreasingly important…esp for your cloud entry project. Cloud is not managed hosting. Self-service cloud is different from managed cloud. AWS is totally different from RAX from a human services perspective. Look in the mirror. What do you need to be successful? Does it make sense to start w/ a managed approach and ween yourself off to self-service?