eWeekChat

A6: Machine Learning has been democratized well across clouds but larger AI is not – capabilities are very different. Edge is also a big usecase for us– there is a strong need for being very close to the customer for latency – driver here “being global and local” at the same time

A6: i see those as big drivers. higher performance edge applications like AI inferencing and retraining are driving the need for multi-cloud support across geos

A6:very much so, in the security space we see higher need for AI solutions that leverage user behavior across multicloud applications to determine risk and take action if needed

A6. Oh related but also different. I think use-cases like industry 4.0, autonomous capabilities, gaming, drones etc do make legitimate edge use-cases. However there's also the edge or AI-washing - things that make you go "is that really needed ?"

A6: Edge (Cloud) is an augment to the Cloud , it is easy to start there and functions like CDN, WAF, BOT, DDOS, real-Time Analytics , multi-cloud observability can give instant benefits without moving to hosting the compute and data in other clouds

A6: Well AI certainly--different clouds are specializing in different types from a service perspective, so as different clouds create different competencies we will see this grow. This also drives data mobility needs. Edge will need to feed data to the right cloud

@GaneshJKRam +1. Interesting that many of the edge use-cases are really about latency and proximity.

@ilakshmis totally agree on edge being part of Cloud leveraging optimization tools to improve experience

A6: we see customers using Cloud data/ML for training large data sets and bring the models at edge for use cases described above, autonomous cars, , Personalized in-store , in-stadium experiences, etc.

A6:  Leveraging #AI-powered intelligence, #ML or #analytics across multiple clouds is important. With increasingly complex operational needs of modern environments, and increasing volume of data - #AI/ML is realistically the only cost-effective way to tackle that.

@arnlopez - security is clearly an area where we see a lot of AI based solutions across clouds. Again, some clouds have far better offerings than the others.

@iLakshmiS Good ones - these make legit cases for edge. Gaming, industry 4.0 manufacturing use-cases are increasingly getting popular...

A6 AI/ML is clearly growing enormously in enterprise adoption. However, it's one of the areas with the greatest difference in functionality among the big three cloud providers. That provides opportunities to pick and choose for use case and drives multi-cloud.

@bernardgolden But also makes trying to develop an ML app that is portable pretty challenging

Media and Entertainment , that includes basically anything that has interactive experience (gaming, live streaming, user generated contents based platforms , are all using Edge successfully. Security and Privacy is also being enforced at Edge for all industries.

@ramsba Heh. Edgewashing. Nice.

@bernardgolden Agree.. each cloud provider is on a journey of their own and ML works as best as the data the cloud provider is subscribed to or allowed to use. So, the value the Enterprise gets is also different from each cloud.

A6 Edge can be more or less like a hyperscale environment. If it is an air gap environment, that can be more challenging in terms of how to run a control plane.

AI and ML capabilities of the cloud are attractive but the required models are data hungry and will bind customers to the cloud provider indefinitely. Multi-cloud portability to have pricing leverage can be absolutely fundamental.

A5: (1/2) Choose tools that work across multiple clouds. If you are using a single native cloud, the provider tools may suffice but the reality is most enterprise customers (1) don’t want to be tied to a single cloud (2) may need capabilities that require multiple cloud platforms

A5: lots and lots of internal discovery and assessment, you need to know where all of your data is, what applications are business critical, do this first, then document and deploy your MC platforms

A5: Cross cloud portability needs to be a design consideration for any greenfield application & part of the test plan. Make opensource or cloud-native technology choices that are standards based & not proprietary ones wherever possible. Use a common DevOps toolchain across clouds

A5. Unfortunately there isn't a thing called "right blueprint" for multi-cloud. Too many nuances and considerations (tech stack, latency, operational tools, skill-set) ... and each ends up creating their own version.

A5: Be consistent where you can. application delivery pipelines, languages, deployment mechanisms. Consistency that still allows for taking advantage of that cloud choice benefits and not abstracting away it's value.

A5: understand what your workload needs are, prioritize your pain points, know the strengths/offerings of clouds under consideration

A5: (2/2) Also the reality is that most #enterprises are hybrid and will have some assets and systems in their data centers and others in #cloud or #SaaS. Tooling from discovery to observability to optimization should ideally support that hybrid scenario.

That said - start small with a few workloads that need to operate across clouds, decide if they are independent parallel clouds or you need interaction. If latter, ensure the experience is good before scaling out. Cloud elasticity is great and all but if you don't start correctly

A5: dont add security post deployment, include this in your overall deployment strategy, integrate security into your multicloud strategy with stakeholder buy in

it becomes unmanageable fairly quickly

@arnlopez - well said. Need to understand all the assets and identify what needs to be multicloud and why so?

@arnlopez True that. I wished multiple elements of infrastructure (security, networking etc) were shifted left in the decision making process

agree with the comments about consistency, making it as seamless as possible by avoiding unnecessary differences between offerings

@arnlopez Agreed. Both security and observability need to be built-in upfront.

multi-Cloud may not be for all, as it may be complex for you to manage, so start with One Cloud, identify why you need it , see the gains, then move to multi-cloud only if you do need.

@ramsba great point on starting out small and building from there. don't boil the ocean

@GaneshJKRam Agreed as well. And discover those assets and dependencies as well as map them to business services.

if you do not have Dev Teams and Cloud Practitioners in house, get trained before you start . You can start with Edge Cloud for CDN, Load Balancer, WAF , BOT etc. So, you can start by leveraging these acceleration and security and data and compute from centralized clouds.

A6 Examine the complete app architecture prior to beginning building to ensure dependencies on a single cloud are identified and portability requirements to other clouds are defined and designed. Post-implementation redesign is expensive and error-prone.

@bernardgolden The challenge with this is convincing decision makers to adopt an approach that triages apps to find candidates where some degree of modernization is warranted versus lift-n-shift only. Hard to optimize spend or do cost takeout w/out touching the app architecture.

A4. (1) Be clear on the architectural choices driving workload placement in single vs multiple clouds. (2) Invest in skillset, tooling, operations - DevNetOps, DevSecOps etc.

A4:
Observability tools can help with managing usage and cost, also need a cloud agnostic security solution to keep your users and data protected

(edited)

A4: Optimize your spend vs. your actual need so that you can proactively avoid unnecessary costs by understanding organic growth trends and future planned events. Like insurance or cell phone plan, it’s important to pick a #cloud provider that is right for your needs & plan ahead

A4: We have leveraged Kubernetes operators that could run in any k8s environment – using cloud agnostic open source or other enterprise software will be critical to success in multi-cloud. There are cross cloud cost or configuration management solutions in the market today.

A4: Ask yourself: "what is my business trying to achieve?" "what are my risks to that result?" if you can answer those it can push towards a decision. Is the main risk time? money? complexity? politics? Work the risks backward

Interesting quote I heard from a recent Enterprise on multi-cloud networking. "I don't want to gold plate my infrastructure and later find that my apps don't deliver". "I'd rather start with the most problematic apps, and then build infra underneath"

A4: Ultimately it comes down to making sure you can deliver service levels customers expect without going over budgeted spend.  Proactive scenario planning to understand how business driver growth can impact your IT resources and spend is important

@GaneshJKRam agreed on the Kubernetes talent, very useful, especially if they know security too

@ramsba agree, workload considerations should be top of mind. this isn't an all or nothing approach

A4 Understand the use case you are trying to address. In our work, we've identified 11 different use cases that all use the term 'multi-cloud.' Each providers different benefits and imposes different requirements and constraints. Use case drives everything else.

@ramsba agree on DevOps, DEvSecOps tooling and making it simple to build, deploy , debug/monitor the apps.

@ramsba - Interesting thought. Goldplating is definitely bad - in the cloud can be cost prohibitive and unnecessary architectural complexities.

@bernardgolden Nice one. In that same order. Use-case > architectural choice

@bernardgolden - good one. Go multicloud only if you need to.

@bernardgolden @ramsba So that means we have to PLAN our deployment ahead of time???

Q4: Make it easy to deploy applications, use standard practices like Terraform providers, automation frameworks , and Multi-Cloud Monitoring and Observability tools.

@bernardgolden +1 for "Start with the outcomes". However, these need to be actually business-centric outcomes and not just technology goals dressed up to pretend like they are biz goals.

@mdcarlson Heh

Q4: Offer specific use cases with some Code samples , API snippets, that work best for those use cases and a way to deploy those templates/code and APIs integration.

Open source and open standards, ideally end-user driven communities. Portability. Requiring app teams to have an data extraction strategy from day one. Don't be afraid to bring cloud concept to on-prem where it makes sense.