Cloud Computing in 2012: What's Already Happening (original) (raw)
Just like model years for new cars, annual technology predictions are arriving earlier and earlier. For cloud computing, the predictions for 2012 started roIling in during October -- before Halloween, perhaps to make them extra scary! The rate of new predictions seemed to peak toward the end of November and compilations of predictions from multiple sources was a popular way to showcase them during December. A few new predictions are still rolling in... after all, it is just now the beginning of 2012.
Instead of adding to the list of predictions, this blog touches on several exciting developments that are already happening. Some of them will even shape the events that are yet to unfold during 2012. Interestingly, some of them are also in direct conflict with several of the common cloud computing predictions for 2012.
- Cloud infrastructure services are further diversifying. While some have predicted an overall commoditization of IaaS, the current reality is diversification. Some providers are copying the AWS API, but certainly not the entire Amazon model and associated services. Some are getting serious about OpenStack, but even there the focus seems to be on retaining the ability to diversify and differentiate from competitors. Some are suggesting that cloud architectures designed for massively scalable web applications have won the cloud architecture battle, yet plenty of cloud architectures are tuned to existing applications that demand greater infrastructure reliability. Cloud infrastructures and their corresponding services are becoming more diverse, not less.
- Hardware still matters and is adapting to the needs of cloud computing. Some like to say that hardware no longer matters with cloud computing. They suggest that "true" cloud computing hides the infrastructure and that application developers shouldn't have to care about that. But haven't programming languages and operating systems always been at least somewhat about abstracting hardware? Yet somehow there are always applications that depend on intimate knowledge of the underlying hardware. For instance, if you want to tune a database, you might just need to know how much data can be copied to and from disk in a single read. What happens to performance tuning in a cloud environment? Sometimes scale out just isn't the right answer. And when multiple tenants are sharing resources, performance levels can be much harder to guarantee. Solid state drives (SSD) -- and associated I/O subsystems are being adapted to address issues like this in the world of cloud computing. Hardware still matters; even in cloud computing environments.
- SLAs still matter and are adapting to the needs of cloud users. Multi-tenancy has proven a bit sticky for many cloud service providers. Among other things, higher levels of resource sharing can make it harder to predict let alone guarantee service levels. Some people are claiming that cloud computing eliminates the need for meaningful SLAs. They suggest that the application developer maintains control of performance and availability through architecture and design. While many new native cloud applications are indeed taking more control in these areas, enterprise IT organizations are still asking for (and beginning to receive!) SLAs based around user experience parameters.
- Cloud computing environments are moving toward higher levels of heterogeneity, not lower. There are certainly a lot of good things to be said about homogeneous cloud hardware environments (greater bargaining power during purchase, interchangeability of components, streamlined management, etc.). However, there are a number of other elements to consider with respect to homogeneity versus heterogeneity. Consider these... a) customers often demand more than one flavor of virtualization; b) they are increasingly asking for access to virtual AND physical servers; c) they are increasingly asking for public AND hybrid environments; d) they want access to other hardware architectures like SPARC; and they take advantage of a variety of storage options. Cloud computing is not a game of one-size-fits-all.
- Solutions are broadening, tackling more aspects of their relevant lifecycles. We are still in the very early years of cloud computing with many solutions still evolving. Point solutions -- those which tackle a narrow problem, often one aspect of a given lifecycle -- are expanding or being replaced by solutions with broader applicability. Cloud management is no longer just about packaging applications and pushing them to the cloud. Cloud management now includes monitoring, performance management, capacity management and change management. Within the application development lifecycle, there are solutions that address software development, testing, deployment, runtime services and more. Cloud enabling solutions are expanding to fill most every void.
This is of course not a comprehensive list of all the great things that are already happing in cloud computing in 2012. I'm interested in hearing your thoughts on other exciting progress -- particularly in areas that conflict with the predictions!!