Monday, July 06, 2015
The adaptive and intelligent datacenter !
It was a big day for the company, every function had planned for many months. The last time around which was actually the first time, the company had fallen flat amidst a huge uproar by customers who were left dissatisfied and unfulfilled. Learning from the fiasco, they had planned carefully looking at each and every component distributed among the three streams which had to work together for collective success. This was their last opportunity to redeem themselves; the team was on tenterhooks, they badly wanted success.
D-day arrived; there was complete harmony between them as if they had telepathic connectivity; intelligence of the highest possible order ensuring that when one requested a resource, it was made available within fraction of a second. Together they ensured that no one was starved of any resources with what appeared to be limitless elasticity. The datacenter team had solved the problem of planned and unplanned demand peaks. They had learned to manage high variability and surge, and now could think of life beyond work.
Who comprised the magic team that found the silver bullet and solved the problems that every datacenter head and CIO struggle with ? Triad of the most critical resources for any application to work: network, compute and storage ! With a combination of Automation, Orchestration, Central Governance, and Dynamic Workload Placement, the ultimate dream of a Software Defined Datacenter – that is application aware and ensures that no resources are ever constrained irrespective of workload – had been achieved.
The journey started almost two decades back with Server Virtualization quickly evolving to allow spread from one server to going across multiple boxes within the rack, then data center and finally the Cloud. With unlimited resources running on commodity hardware, it’s now possible to react to sudden and planned upsurges. Cloud providers have built capacity that collectively can address the entire world’s needs on steady state. Economies of scale have been achieved by many; the decision to build dedicated datacenters is now an emotional and not rationale decision.
Software Defined Networks have evolved to allocate resources based on workloads managing capacity to optimize. They are application aware and prioritize traffic based on neural network based learning; the disruption to existing hardware players had them acquiring some of the new innovative startups. Finally hardware independent storage virtualization and automation brings up the rear having evolved last and still facing some challenges of interoperability threatening stranglehold of dominant players. Collectively SDDC has arrived and gaining momentum !
Software Defined Datacenter is a complex architecture of hardware, network, compute, storage, and unified software that is application context aware working within the boundaries of defined SLAs and configured for business. The abstraction layer thus defined, delivers everything as a measurable software configurable service. The business logic layer is intelligent to identify trends and constraints overcoming them in an interconnected world of on premise datacenter and Cloud (compute, storage, and applications).
In contrast the current traditional datacenters focus on discrete components – hardware and devices – which require individual configuration, provisioning and monitoring. Depending on the maturity curve, enterprises have adopted various parts of SDDC in their environment, whereas the mature service providers have come closer towards the creation of the software defined datacenter. Interoperability and transition between competing SDDC propositions still remains a challenge with proprietary and still evolving software stack.
So where should IT and CIOs focus or place their bets ? Can they create the magic described above within their datacenters ? Is it feasible with permissible or marginally higher investments ? Which vendors should they bet on and how to manage the existing stack of hardware and software ? Does it make sense to just outsource the entire proposition and let someone else manage the solution and associated headaches ? What about skills availability to create and then manage the complexity that SDDCs bring ?
I believe that the answer lies with the existing scale and intricacy; SDDC can bring high level of efficiency to large datacenter operations where the variability of demand coupled with dynamic scale provisioning will bring efficiency and potential cost savings over the long run. The other scenarios that make a case are business with unpredictable load and variability (e.g. ecommerce portals, news sites, stock exchanges, and social media). Smaller companies can derive benefit from parts of the solution with high levels of virtualization.
Either way, start building components that can help you move towards SDDC.