Performance & Capacity Planning: Important Issues of the Cloud Computing

Once upon a time, there used to be a roomful of UNIX machines providing computer services to the remote dumb terminals. You get the connectivity via the telephone lines and then you telnet the remote computer that had your account. You used all the software available on that machine. But, the service was limited to the users of the respective organization.

Then, the PCs, INTERNET and the servers associated with them arrived. We started having our own software installed on our PCs. We started using the INTERNET for everything. This also opened up the security issues. Security software, one of the most resource consuming software, became an inherent part of the PC. As the computing powers grew and the memory price dropped, the applications became more and more CPU and memory hungry. As a result, network connectivity got increased in bandwidth. So, we started doing a lot more stuff using the network. Soon we realized that it would be nice if we can have all our applications, which have also become very expensive compared to the hardware, provided by the service providers based on a on-demand basis . Slowly, we entered the cloud computing (CC) era, which almost looks like going back to where we came from, but in a humongous scale and with a different paradigm.

Apparently, CC seems to be the next step in the INTERNET evolution. Cloud computing, as the name implies, is a concept where the computational capability is provided in the cloud or the internet as a utility. The computational capability may include Software as a Service (SAAS), Hardware as a Service (HAAS), infrastructure as a service (IAAS). There are great many advantages with the CC [1]. There are already many major infrastructure providers or Cloud Computing Service Providers (CCSP) in this arena, such as GOOGLE, AMAZON EC2, etc.

While the name CC sounds very exciting, and there is a lot of hype about it in the IT world, it is still very hazy to many experts as to what exactly it is supposed to be and how it could be provided. Many people interpret it the way their background allows them to understand, which reminds us of the 'Big Elephant and the Seven Blind people'. However, the current emphasis in the CC seems to be mainly focused on the application software side. It is true that many people recognize the security and reliability (or is it availability?) issues as well. But, my search for papers on the performance analysis and capacity planning of the computing resources, including the network bandwidth in the CC infrastructure, did not yield any useful results. Neither did I come across any paper that would suggest any optimization of the system as a whole. The closest one is the Berkley paper [2].

In [3], Bernard Golden rightly recognizes the topic of bandwidth capacity in the CC as a very crucial issue. He calls for modeling and simulation related to the traffic pattern to predict and estimate the capacity requirements of the links to the external world from the cloud. It should be noted that the capacity estimation should be considered even within the cloud as well as the capacity requirements of the links that connect the cloud to the users or cloud clients (CCL).

Many of the issues that are to be addressed in the CC are still in the clouds: 1) How to provide SLA : Reliability, Availability, Bandwidth Guarantees, QoS, Security, etc., 2) How to define QoS , 3) How to do the capacity planning and traffic engineering for the intra and inter clouds, 4) The performance analysis and modeling of the clouds, 5) Systems engineering the cloud to build an optimal cloud (optimal selection and distribution of the data centers with optimal connectivity between them) and optimizing the resources. And, for the cloud user or CCL, how and when to choose the cloud service provider such that the profit is maximized. All these call for the need for the right systems engineering teams by the CCSP and CCL.


References

1. 10 Things You Don't Need to Do In the Clouds, http://smoothspan.wordpress.com/2009/05/25/10-things-you-dont-need-to-do-in-the-clouds/
   
2. Above the Clouds: A Berkeley View of Cloud Computing, http://radlab.cs.berkeley.edu/
   
3. Bernard Golden, The Skinny Straw: Cloud Computing's Bottleneck and How to Address It,http://www.cio.com/article/499137/The_Skinny_Straw_Cloud_Computing_s_Bottleneck_and_How_to_Address_It