This white paper performs a comparative analysis of the two concepts governing the present and the future, respectively, of the IT industry: the older, already-proven technology used in Clustering systems today, and the new trend of Cloud Computing; with a special focus on the development of email services.
Lately, there has been a lot of buzz about the brand new cloud computing paradigm. Supposedly, through this type of cloud resource, you would get the same performance and user experience levels as with classic systems. Some companies have moved word and spreadsheet processors online, while others plan to allow users to play HD quality games on demand, over the Internet – all striving to move from regular desktop related applications onto the Web.
Cloud computing yields a series of palpable benefits such as a lower TCO for both service providers and their customers, while also attempting to overcome the main concerns of privacy and security advocates in the IT industry.
This white paper aims to compare the older, industry-proven technology used in clustering systems today and the new trend in information technology – cloud computing, and to answer some of the questions revolving around these topics.
Will the Cloud Replace the Cluster?
This is surely the most obvious question of all when it comes down to analyzing the future development of the situation. However, it is one of the most interesting as well, because cloud-type resources are actually served by what is called a decentralized cluster.
Thus, while on a regular cluster you have a central nervous point (the storage, for the highly available one, or the controller for grids etc.), in a cloud all systems perform any and all tasks just as good. The cluster is a well structured and divided environment that relies on role configuration for each node and workload distribution according to technical limitations. The cloud, on the other hand, will divide the workload internally according to external requirements and user preference, rather than technical limitations.
Within a fully developed and featured cloud, the concepts of statefull and stateless process / information no longer have any impact on how the system reacts. All data is readily accessible to any computer in the setup; moreover, any computer can serve any request initiated by any client. This extensive use of “any” is what makes the cloud so wonderful to the industry in the sense that it offers a lot of freedom and that will most likely re-write most of our preconceived ideas about how communications take place.
However, node roles may still exist within today’s cloud platforms, as with regular clusters (i.e. front-end, back-end, mailstore, database etc.). This is part of a natural evolution of the cluster, very much like a transition state. Consequently, the cloud will develop into this magnificent entity later on, as we learn to deal with the burdens of deploying one along the way. This will essentially be the natural evolution of the cluster and the only direction that makes sense given the technological status and demand of the industry.
How Do the Two Technical Solutions Compare?
This second most likely question is very adequate as well: why would anyone need to replace a cluster in the first place? Such high-end solutions are expensive and very resilient on their own, so is there a real need to go any further?
To answer this, we’ll have to take a look at some widely used systems to see what’s currently missing and what else can be gained by such a move.
We’ll use email systems as an example because each and every company today uses either an in-house (in-sourced) email server or a hosted paid service run by a third party (outsourced). There is no other way around email communication services, short of using the FAX system to deliver documents.
Email clusters rely on two or more layers of systems (usually called tiers) that perform tasks within one of the following categories:
* Statefull tasks – they rely on persistent data that must be saved for each session, like the contents of an email message.
* Stateless tasks – they rely on temporary information that is relevant to one individual session only, such as the authentication state of a connection.
As stated before, such categories do not apply when deploying a cloud setup, as all systems are able to perform both statefull and stateless tasks. This fact results in significant advantages such as the full balancing capability between the systems. This specific capability is also inherited from the cluster technology, by the AXIGEN Mail Server and several of its market competitors. However, all products are limited – as far as regular clusters are concerned, because load balancing is not possible in the case of statefull transactions; using a cloud makes this possible and solves some of the most important puzzles related to email clusters today.