Making Clouds Carrier-Grade
Cloud computing is increasingly being used for applications that have high availability and performance demands. Telecommunications, with its infamous 'Carrier-Grade' tag for describing such requirements, is one example. To facilitate the dialog between applications and cloud providers, the Asia Cloud Computing Association (ACCA) is developing an Index, that defines Carrier-Grade in a cloud computing context, in a structured tabular form. This establishes a normalized baseline and language with which to discuss needs and capabilities. By doing so, the ACCA hopes to stimulate usage of clouds for demanding applications, like telecom, and others as well.
ARE CLOUDS RELIABLE ENOUGH?
As global operators start to use cloud computing solutions for both internal and external services, the question that arises is: 'Are clouds reliable enough?'
Certainly, by virtue of existence and usage, we can say cloud computing is 'good enough' for a very large and growing number of service types. However, cloud based telephony services bring new demands. The infamous five 9's reliability that has been the hallmark of telephony cannot be achieved easily, if at all, using typical cloud computing hardware and software systems.
Why? The fundamental principles of implementation are different. Telephony often relies on highly customized, costly redundant hardware implementations combined with sophisticated software know-how to meet the demanding five 9's requirement. Cloud computing typically takes an opposite approach and uses off-the-shelf hardware, without redundancy, to save costs, and relies almost entirely on applications to implement recovery strategies when failures inevitably occur. The shift in ownership of reliability is significant which raises the question: Can such systems ever meet the resilience needed by telephony?
SOFTWARE RECOVERY VS. HARDWARE RECOVERY
The answer of course, is, it depends. While the discussion has so far been about telephony, one can certainly imagine other industries having similar needs. Banking and stock trading come to mind, as do government, military and health services. In practice, each industry has unique demands to be met.
But as the economics and simplicity of cloud implementations start to become overwhelmingly attractive, it is natural to think that the demand for varying grades of service from cloud providers will come. One size (as the cloud computing largely is today), will not fit all. Looking to get mass telephony on your cloud and attract all the big operators? They you had better have a story that makes implementing five 9's availability easier than on the competitor's cloud. Trying to attract Skype as a customer? You might want to make sure you're round trip delay is less than the expected 200ms needed for good quality voice.
WHAT PARAMETERS MATTER
But what are the right parameters? 'Carrier-Grade' is defined by a large number of complex requirements of which five 9's availability is only one. For example, latency, the reliability of links to a cloud, and the availability of 24/7 support, are all components of a carrier-grade system. It is a daunting task even for a technical expert to know the right questions to ask and the relevant standards that apply. An 'Index' could help, but who should do it? An independent body is likely the best suited to avoid bias.
THE ROLE OF THE ASIA CLOUD COMPUTING ASSOCIATION
The Asia Cloud Computing Association is doing exactly that. Within the Association's Carrier Grade Cloud Computing Working Group, we've identified a table or "Index" of relevant parameters and their values that that define the typical needs of high performance applications. That Index can then be used by application developers to specify their needs to cloud providers. It can also be used by cloud providers to specify the grades of service offered by them. By using a standard Index, the dialog between the two parties is and a consequently, a match of needs to offer is simplified.
Imagine the alternative. Without a standard Index, each party will use a different set of parameters, definitions of those parameters. The result is a complex matching process with a high chance of mistakes being made. For very large corporations, this may not be an issue, where resources can be brought to bear on the topic, but for smaller companies, it may simply be too over-whelming and a more conventional, non-cloud solution might be pursued.
All that being said, we recognize that for some applications, like mass telephony, some needs may never be met on public clouds (like five 9's availability), but if the delta between the need and the offer are clear, then the application implementation requirements to overcome any gaps also becomes clear. Clarity in this regard is essential.
So what do we hope to accomplish with the Carrier-Grade Index? The objective is to stimulate the industry to use cloud computing for applications that may not have been previously considered - by making it easier to match application needs to cloud offers. We believe an unbiased group will best accomplish this, using standard definitions of requirements where possible, in a tabular form, i.e., the Carrier-Grade Index.
IS THIS RELEVANT?
In private discussions with a leading operator, it was revealed that the first step in moving to cloud computing for them was dominated by a drive to reduce costs for internal systems. In doing so, the demand for 'Carrier-Grade' became apparent (from internal customers), and as a result, their own cloud computing implementation had to consider multiple tiers or grades. It does not take much imagination to consider this being a common theme in the telecom industry, and others. And if Google search can be considered a measure of interest, "Carrier Grade Cloud Computing" resulted in 220,000 hits on Jan.15th, 2012! For these reasons and others, The Asia Cloud Computing Association believe this to be an important topic and plans to help advance understanding of it, and the Carrier Grade Cloud Computing Index is one of those ways.
ABOUT THE AUTHOR:
Michael Murphy – Asia Cloud Computing Association Board Member
Michael Murphy is Head of Technology for the Asia-Pacific region of Nokia Siemens Networks supporting customer engagement activities related to new technologies.
Mike has an extensive background in Research & Development and has lead both business and R&D teams in China, Korea, Japan, Thailand, Turkey, France, and Canada.
Mike has a Master's Degree in Mathematics from the University of Waterloo, in Canada, and is now based in Bangkok, Thailand.