2009-09-10 02:09 PM
You can take en easy way & a hard way:
Easy way - a company could e.g. use Google for their email (external cloud) & have their own infrastructure for other apps, e.g. file shares (internal cloud)
Hard way, where the question becomes really interesting - how to move the same app / workload between private & external cloud? IMHO there is no good answer to this one at the moment, but e.g. VMware is claiming they are working on some standards to make this magic doable.
2009-09-12 05:20 AM
That's an interesting question. From my perspective, I expect to see many companies build or leverage tools and applications that will allow them to monitor and manage business services that exist in a private or public cloud. These will come in the form of orchestration tools that abstract all of the business app, compute, network and storage pieces that make up a specific business service.
I don't expect to see business services span both the private and public clouds in the immediate future, but it is an intriguing option. In my opinion, you'll see companies leverage private clouds for business services like nuclear plant random sampling apps or transaction processing apps that contain protected data like credit card or SSN they want to hold close to home and have complete control. They'll leverage public clouds for those business services like e-mail that they are okay with letting someone else manage.
What types of business services do you see that would span both the private and public cloud?
Cloud, Database, and Business Apps Global Field Technology Lead
2009-09-19 08:14 AM
Hi Bill! I agree with you that in the near future that companies will keep their critical businesses in house. However, I see a gradual progression in the comfort level of spanning critical business service across both the internal and the private cloud. This progression will happen at two levels:
This is very similar to the adoption curve of virtualization technologies and their prominence in dev/test environments. The idea is to get new systems up and going. Once done retire or re-use at a later time without having to maintain "inventory". We will probably see the same concept stretched to the external cloud
Key business service such as the Insurance under writing, online banking, Supply Chain etc are composite applications that have multiple parts/components and span from Mainframe to Mid tier, Load balancers, DBs etc. Companies might be more comfortable in stretching parts of the service to the cloud and not the complete application
The key influencers to the adoption are probably going to be following:
This might be easier said than done as customers do not have good internal metrics to keep track of their own costs to benchmark against what an external service provider offers. (This opens up a different discussion on how Service Providers position and market their offerings to overcome this). In my opinion, cost transparency is a better influencer of behavior as opposed to charge-back as company dynamics differ with regards to charge-back but are consistent when it comes to cost transparency.
Marketing campaigns such as dollar a day per VM are innovative, but the key is going to be around how fast and relialbly new services are provisioned in the environment. And do they provide the required service levels? Internal IT always things that they can do the same job better and quicker. This in many cases is for right reasons.So the cynics will always be out there.
It will start with components that have the least security and compliance regulations We probably need to adjust how we audit and run internal compliance on systems that span both the private and external cloud. A side question is - How do you effectively secure the data that is being resides in the internal cloud but is being accessed from a system in the external cloud?
Shifting expenses from CAPEX to OPEX has advantages but the challenge is around how do you effectively balance the two?
I have not put a lot of thought around this but this is something that the Lines of Business need to account for. Budgets need to be predictable and companies need to do a balancing act between what shows up as an Asset and what shows up as pure operational expense. I would love to get a broader insight about this from other members of the forum
Some thoughts here. Would love to get additional comments etc on this.
2009-09-23 11:32 AM
You raise some interesting use cases. I agree with many of them. Starting with Dev/Test I can see where you would think what's been going on with virtualization would be a good indicator as to what might happen with Cloud technologies, but I would still feel pretty strongly that you won't see data that has been deemed to require a private cloud environment being allowed to enter a public cloud space. True, the cost model should entice companies to look at this as an option, but they will first have to wrestle with the idea of where they are willing to let their data reside and their comfort level with that decision. There are some great technologies that will enable customers to alter their structured data for these purposes, but there will still be the concern for the application config data that can't really be altered. This will definitely be something I'll be watching closely as the NetApp Cloud, Database and Business Apps Field Technology Lead.
As an aside, if you're not familiar with our App Dev-Test solution, please visit this URL: http://www.netapp.com/us/solutions/infrastructure/app-dev-test/. I was one of the primary contributors to the development of this solution. I think you'll find our value proposition in this space very interesting.
I truly agree with your second use case. I see business services being categorized and even parts of the service being further "tagged" as candidates for public Cloud, private Cloud, virtual compute, and physical compute infrastructure. These tags would not necessarily be what the end business consumer sees when they purchase or consume the service, but more of the back-end compute/network/storage infrastructure supporting the service. My assumption is that there will be a variety of combinations of these pieces that will comprise SLAs and drive the cost model.
Speaking of cost models, I agree that this will be the primary influencer of adoption of both private and public clouds. Time will tell as public cloud providers become well established whether or not the out-source/in-source decision will be an easy one based on cost. I believe that most companies that are heavily compute virtualized will continue down that path and build out a private cloud infrastructure.
Whether talking about public or private cloud, the other influencers you note (performance, reliability, compliance and security) are also very valid. My thoughts on these are that each one will fit into a defined SLA. Some, like security and reliability, will be required across the board. Others such as performance can have varying levels that might be the differentiating factor when defining one SLA versus another.
Finally, I'm of the opinion that the CAPEX vs OPEX question is something that will vary based on accounting and capital needs of a given organization. I too, would love to hear from others on this topic.
Thank you for your thoughts on this topic!