Takeaways from Carrier Cloud Forum @ Interop 2011

Today's takeaways from Interop 2011--Carrier Cloud Forum panel on "Building the On-Demand Cloud Infrastructure".

• Private cloud, by definition, is not "infinitely" scalable like Amazon's EC2. The resources required to build such a solution would be prohibitive. In addition, private clouds are commonly being delivered and managed by service providers. Because of these characteristics, there are some that claim that private cloud is not truly cloud computing. Of course, managed services have been around a long time and there are still efficiencies to be had for customers who take advantage of carriers' economies of scale and scope.
• Why service providers, why cloud?
• Billing is crucial: unified billing is a major advantage for carriers who offer end to end services; pipe--DC--cloud in a one-throat-to-choke model. This also allows carriers to provide contractual SLAs on the full service offering.
• Carriers already have a critical mass of customers who can benefit from this bundling as well.
• Cloud computing depends on the network working and carriers own the network.
• Carriers are looking for a partner who can provide a solution that is competitive with the big technology vendors; as robust an offering but with none of the drawbacks of a large organization.
• Cloud computing is pervasive: it is difficult to see where it starts (i.e., define its boundaries) as it encompasses mobile, web, business applications, enterprise 2.0, social media...
• What does "enterprise grade" or "carrier grade" really mean? IaaS is usually referred to as being built on servers that do not have redundancy inherently built into them (e.g., single power source, non-raided drives, etc.). Perhaps its better to drop the "enterprise grade" and simply refer to it as "carrier grade" since carriers have historically been concerned with providing redundant services (such as telephony, internetworking) the logic being that carrier customers want the reassurance that the hardware will be tolerant to failures.

Misc. Tidbits from Interop 2011

This week, at Interop, I attended Private and Public Cloud Days in which speakers discussed the relative merits, drawbacks, and case studies. I've compiled a few ideas that I thought were interesting:

• Regarding private clouds, we have commonly referred to them in the context of dev/test, but in reality, their value is much more important as a way to deliver "IT as a service".
• Since clouds are commonly built on hardware that is not designed with redundancy in mind, cloud computing does not automatically imply redundancy. High availability is included in the architecture of your instance footprint (e.g., no IP hardcoding, no DB source hardcoding, etc.) and application(s) running on it.
• Redundancy costs money. Because of this, your organization needs to understand its tolerance to risk (and recovery point and time objectives (RPO/RTO)) and then design to meet this tolerance level.
• "Big data" should be actionable. Instead of thinking in terms of data, think in terms of business problems and solving those.
• Fear of lock-in from platforms (PaaS) is irrational since programmers have been locking themselves in by the simple act of choosing a programming language.