The technology ecosystem that we all interact with continues to move and shift with an ever-increasing pace. The digital era, the information age, the connected age, the social web and cloud computing all revolve and evolve around us.
Whatever the current label or buzzword is, it’s clear that we live in an information-intensive world. Consumers of services expect their journey to their desired outcome to be simple, accurate and timely. The profitability and competiveness of businesses increasingly depends upon their ability to rapidly extract value from their information assets.
The traditional data centre stares into the abyss of our information-intensive world. Outlining the scope of the challenges we face, IDC predicts that by 2020 the amount of digital data will grow 44 times to 35 billion terabytes. But the aforementioned trends do not just increase the amount of digital data stored. New mobile technologies allow users to access this data from anywhere at any time through multiple devices, significantly increasing data traffic. In fact, IMS estimates that by 2020 there will be 22 billion independent devices connected to the internet – near 4.5 times more than 2010. And our patience for accessing information is growing shorter every day – enterprises and consumers expect information systems to be real time, on-demand, always-on and highly responsive with relevant information.
Unless our IT visionaries take hold of opportunities offered by this need for change the information abyss is only going to get deeper and darker. How could the next generation data centre address these challenges?
Next generation applications and application tuning will certainly play a role – but how significant could their impact be on their own? As the richness and complexity of data processing grows, the working dataset sizes are beyond what can be achieved in DRAM either commercially or due to power density limitations.
Server compute power continues to push forward and seems to offer hope to the next generation data centre. And yet, as soon as the working dataset is beyond the grasp of DRAM - held solely in the centralised storage systems that have embodied many of the data centres of today - the CPU is idle held to ransom by I/O wait. Each year, the gap between the rate at which a server can process data and the rate at which legacy storage devices can deliver that information widens, and the resulting data supply problem holding back the CPU becomes tighter.
The empirical effects of underutilisation of CPU seem clear: IDC report that in 2009, over 80% of servers were idle half the time and 37% of servers were idle 80% of the time.
Given this, squeezing as much compute power as possible out of the CPU will become the imperative as businesses fight to create an edge and control costs. How can this goal be achieved? Already today, thousands of enterprise customers have been accelerating their ability to create information value or reach their end goal by placing NAND flash memory in the server, close to the CPU where the data is processed.
With PCIe-based NAND flash technologies, customers can either manually or automatically place their active, “hot” data sets in non-volatile memory inside of the server next to the CPU. This allows customers to stretch the performance and capacity of the solutions they have already invested in by serving the hot data from the memory in the server, and holding the “colder” data on the centralised storage. By accessing the power of the CPU, rather than using an embedded controller, these PCIe solutions can evolve in tandem with the powerful processors running the server itself. This can help eliminate the data supply problem burdening our data centres.
Many companies adopting these approaches have reported achieving greater than 10 times the application throughput per server through increased server utilisation. This increased performance is often used as a mechanism to facilitate real data centre consolidation, resulting in reductions in ongoing facility, energy and cooling expenses.
If we don’t embrace true innovation and change the way we think about data centres, the abyss won’t go away. It’s time for true leaders in IT to guide us in approaching the problem differently. It’s time to embrace data centre evolution.