Data centres run our world, managing petabytes of information as it is networked through a labyrinth of wires, switches, servers and adapters. Some data centres are large, others small, some are built upon disk, whilst others are moving towards all flash architectures.
A few years ago, industry experts said the all flash data centre would never materialise. With all of the biggest data centre hardware vendors now providing flash memory solutions, companies are implementing flash at ever increasing rates. While no one is predicting the demise of disk, flash is no longer used by only early adopters or large hyperscale companies. Flash is becoming the standard in the server and moving to other portions of the data centre, continuously being used in unique ways to accelerate new applications.
But what is driving the evolution of the all flash data centre?
The need for flexibility and performance
More flexible than disk, flash provides capabilities that are not possible with disk era architectures. In many ways, this flexibility enables IT professionals to accelerate many portions of the data centre with intelligent software rather than relying on costly hardware. When data centres make this kind of flexibility a priority, IT managers will achieve higher performance at lower costs.
As data centres continue to evolve, performance is an obvious catalyst for change. Thanks to advances in architecture throughout the stack, applications are much more powerful than they were five years ago. Flash memory in data centre servers are a key component to unlocking the full performance potential of today’s powerful systems.
The performance benefits of flash are not only being observed and implemented by giant technology-driven multinationals. After experiencing delays in turning around credit reports, credit reference agency Callcredit Information Group made the switch to flash and improved performance density by more than 14x. This meant customers could get faster turnaround times on their credit reports, greatly improving the value of Callcredit’s services.
All Assets at Work
CPU performance improvement rates are continuing to hold true today in accordance with Moore’s Law, just as they have since the 1960s. But without sufficient performance, today’s powerful CPUs are now becoming more underutilised. In fact, today, due to many enterprise architectures, the average CPU is only 10 percent utilised because they are not fed data fast enough.
With the help of flash, and all flash data centre implementations, IT managers can better utilise more of the performance capabilities of the CPU, enabling businesses and enterprises to handle more data with less hardware in their data centres.
By placing flash memory in servers where they are near the CPU, applications can better utilise processing power by eliminating network bottlenecks. Using 100 percent of the CPU could be analogous to people unlocking new abilities by using 100 percent of their brain power, rather than the traditional 10 percent. Flash memory is helping make this possible in the enterprise, powering new applications that are changing our world.
Cash is King
While staying in the know about game-changing breakthroughs like flash is a priority for every IT manager, one roadblock that sometimes halts actual implementation is cost. Even as IT budgets increase, cost and pricing continues to play a significant role in data centre upgrades.
Today, many flash memory implementations are more affordable than disk-based implementations, especially when considered from the total cost of opportunity perspective. Flash helps improve performance density, reduce licensing costs, and prolong the life of existing architectures. Many only compare dollars-per-gigabyte when comparing disk and flash. Real value is uncovered when comparing dollar-per-performance. A disk drive maxes out around 210 IOPS, however flash drives are capable of achieving between 500,000 and 1,000,000 IOPS.
In addition, flash is available at very affordable price points. With companies like Facebook embracing open architectures, and sharing its designs through projects like Open Compute, companies of all sizes can take advantage of the latest concepts in data centre architecture.
Winning on Energy
Looking at costs, it’s important to note that it’s no secret that data centres consume vast amounts of energy. In fact, it’s estimated U.S. data centres consume 61billion kilowatt hours (kWh) a year - enough to power a home for 6.4 million years. Half of that energy is consumed by data centre cooling infrastructure that keeps hardware from overheating.
There has been immense call to action for data centres to be greener, consume less energy and reduce their carbon footprint. In fact, a study released by the Environmental Protection Agency (EPA) shows that US data centres could reduce energy consumption by up to 70 percent using new technologies like flash memory.
Because flash memory consumes less energy, it also requires less energy to cool, making it a great fit for the environmentally-friendly data centre. Once flash is added into the data centre, improved performance from flash allows IT managers to reduce the amount of energy-hungry DRAM needed in their system and increase content served, all while lowering energy consumption.
Legacy data centres are big, inefficient and expensive and they consume vast amounts of energy. Whether you’re hoping to shrink the size of your data centre, reduce overall costs, improve performance or save the environment, flash can help deliver the capabilities you need.
While you might not be ready to go all flash, there has never been a better time to begin to incorporate flash into your IT infrastructure. As the benefits of flash powered data centres continue to grow, flash is becoming an industry standard, so testing the waters now will ultimately pave the way for the data centre of tomorrow.
Posted by Mat Young, Senior Director of Data Propulsion Lab at Fusion-io.