Data centre automation is no longer the preserve of hyperscalers
Any organisation facing constrained budgets and resources, while trying to deliver a hybrid cloud service offering, should now be looking at data centre automation. It’s not “whether to adopt it”, but “how far to take it”
If you want a taste of what network automation can achieve, take a look at the data centres powering today’s Super 7 – Amazon, Facebook, Google, Microsoft, Baidu, Alibaba, and Tencent.
This year’s Economist feature on competition in the digital age talked about the sheer market presence and technological dominance of these “tech titans”, and their threat to established leaders, particularly in the retail sector. When we factor in second-tier “hyperscalers”, like LinkedIn, Twitter and Instagram, we can estimate there to be around 500 next-generation data centres worldwide.
What has changed is that the specialist providers that originally emerged to serve the hyperscale market are now looking for a bigger market opportunity
Those companies do business on such a massive scale that they have been able to afford teams of hundreds or thousands of highly qualified engineers to develop their data centres.
In-house they needed to create something that was not otherwise available – a new generation of data centres, offering unprecedented power to mine data and automate business processes for super-efficiency.
It would take years and countless man-hours to achieve, but hyperscalers – from their dominant position – are willing to spend time to save money. It is different for the average firm up against heavy competition. To keep ahead, CIOs are forced to spend money to save time.
Ultimately, organisations wants to achieve the same thing, but few have the money or resources to follow the hyperscalers’ in-house development cycle, or even to afford the heavy cost of ordering such a system from traditional IT suppliers.
What has changed is that the specialist providers that originally emerged to serve the hyperscale market are now looking for a bigger market opportunity. They want to extend their smart interconnect solutions, specifically designed for performance and minimal latency, to the massive enterprise market.
They are simplifying the deployment and lowering the cost of hyperscaling to bring it within reach of medium to large firms. New companies are also emerging to serve this market, enabling hyperscale automation for the masses.
Automation for the many, not the few
What are these companies offering? The first requirement is performance. A good data centre network is like good plumbing: you don’t want the tap to run dry while someone takes a shower. So hyperscalers threw bandwidth at the system.
Your business might not justify 100G ethernet, but it makes no sense to buy 10G when the price of 25G is almost on a par. A company like Mellanox has been supplying hyperscalers with high-performance ethernet switches, adapters, cables etc and is now offering similar solutions for the enterprise market. Network performance is no longer an issue – the first essential for an automated network.
A second essential is for open systems that prevent a firm being tied to a single supplier. Whether your priority is to use low cost “white box” components, or to be able to select “best of breed” for critical applications – the focus with all these suppliers is to allow equipment from any supplier and to be compatible with all the main software platforms.
For example, Cumulus’s Open Network operating system and networking automation specifically enable webscale networking via white box equipment, without breaking the bank.
New companies are also emerging to serve this market, enabling hyperscale automation for the masses
It used to be a colossal task to build automation into an existing system, with so many components needing painstaking configuration. People think of automation as a way of saving manual work, but it has another, even more important role.
Repetitive manual configuration is the sort of task that leads to human error and system failures that can be very hard to diagnose and correct. Automation eliminates human error, and frees up brain work for more rewarding tasks.
Take Intent Based Networking (IBN) for example. Apstra is a company offering automated network roll-out, configuration and updating from a single graphical interface that also provides visibility into the entire system performance and potential for automatic fault correction.
The system understands an intent such as “connect 500 servers via 10G-leaf via L3 routing, with 1:1 oversubscription”, and automatically generates a blueprint for acceptance, identifies potential problems and finally monitors the performance to make sure the intent is successfully met.
Another key lies in network convergence – supporting compute, communications, and storage on a single integrated fabric.
This allows the system to scale outwards by adding units of CPU, memory and storage independently, rather than having to disrupt the existing structure by building them onto it.
People think of automation as a way of saving manual work, but it has another, even more important role.
Another company, Nutanix, offers Hyper-Converged Infrastructure (HCI) – software that integrates compute, storage, and virtualisation in one easy to deploy and use package.
Hyperscalers had to develop all this automation themselves from scratch. Now, with solutions like these, organisations can get the advantages of webscale without that R&D budget.
The final key is integration. For example: Nutanix effectively makes the infrastructure invisible, while Mellanox now synchronises with the Nutanix hypervisor to make the network invisible too – like the very best plumbing that simply works. The resulting data centre infrastructure becomes exceptionally easy to deploy and manage.
With data centre automation, companies can rapidly and simply achieve hyperscale levels of performance, efficiency, and agility. This is a very significant development. The focus now shifts from specialist network expertise to optimising business applications.
Any firm facing constrained budgets and resources, while trying to deliver a hybrid cloud service offering, should now be looking at data centre automation. It’s not “whether to adopt it”, but “how far to take it”.
ο Kevin Deierling, VP of marketing at Mellanox.
Tags:AI automation data centre hyperscalers hypervisor intent based networking robotics
READ MORE >>
READ MORE >>
READ MORE >>
READ MORE >>
READ MORE >>