Demanding workloads are putting more pressure on IT infrastructure fueled by social and mobile applications and big data analytics. For organizations that run a Linux environment, many have opted for x86 servers in the past, but the change in the nature of applications in the digital age means a re-think of hardware choices that can offer the performance, reliability and security these business-critical applications require.
A report by Forrester, Performance and flexibility serve the enterprise, commissioned by IBM, highlights the considerations IT executives face when selecting servers for Linux installation.
“Companies that hope to stay competitive in the marketplace must begin to consider alternatives to x86, or else they will find themselves at a serious infrastructure disadvantage,” it warns, and advises IT leaders to plan a clear path to a more robust, performance-oriented architecture as applications grow in depth and complexity.
Clive Longbottom, founder of analyst group Quocirca, outlines the key options when it comes to selecting the best servers to support Linux in an IT infrastructure.
“The choices come down to Intel/AMD, IBM Power or a mainframe/LinuxONE, with the option of I/PaaS (infrastructure or platform as a service) public cloud offerings,” he says.
Experience the journey of open development and collaboration, driving technology innovation with Linux on Power Systems.View Now
Question 1: what skills do you have?
The first consideration is to look at what internal capabilities exist within the IT organization.
“What skills do you already have? If it is all around Intel/AMD architecture, then Linux on Power may not fit into your existing skills and tools sets,” says Longbottom.
However, this might not be the over-riding reason for sticking with old habits if other considerations outweigh the questions of skillsets and tools. Organizations do need to be prepared to adapt in order to accommodate the requirements that will be made of their architectures.
“If the skillset doesn’t worry you so much, then maybe it should be more of a question of what sort of scale and flexibility you want,” says Longbottom.
He says that if you only have a few Linux images, then Intel/AMD is probably going to be good enough, but Linux on Power is a great alternative. However, for companies that are anticipating bigger workloads, he says scale is the over-riding factor.
“For bigger scale, Linux on Power and start to make sense,” he says.
Linux on Power is the primary processor for those customers who use Linux, especially those who need compute-intensive workloads, such as big data, analytics and cloud.
Forrester’s report highlights that in the skills versus scale conundrum, there is likely to be a cross-over point.
It gives the example of a healthcare CIO who says that support staffs are locked into the current hardware ecosystem, but that cost risks exist if the organization persists with this architecture because “complexity and layers added for data analytics, cloud integration, and a bloated server footprint are beginning to tax current capabilities and challenge conventional thinking.”
Question 2: What scale will you need?
To focus minds, Quocirca’s Longbottom says the key question is, “What scale of Linux deployment are you looking for?”
Linux on Power is able to scale and Forrester says this is critical as IT professionals that stick with commodity operating environments are beginning to see limits.
“The concerns revolve around applications’ scalability, agility, and ability to combine transactional data with other analytics,” says the Forrester report. It recommends that as complexity and volume of applications and data grow, a more powerful and flexible hardware environment, such as Linux on Power, becomes compelling.
Throwing commodity hardware at a problem will only increase frustration - it is more sensible to adopt a holistic, forward-thinking view, and for some organizations it is clear that x86 doesn’t scale fast enough.
ZTEsoft, a subsidiary of equipment manufacturer ZTE, found that scale was the critical challenge when it wanted to expand into new regions. It tracks customer data interactions to better understand client behavior and in the past ran Linux on x86 systems, but it was willing to consider IBM Power Systems for the scale and performance required for these workloads.
“Analyzing big data requires a powerful, scalable platform,” says Yang Bo, from the ZTEsoft Global Marketing Unit. “With Linux on IBM Power Systems, our customers gain the performance and capacity to analyze a rapidly growing volume and variety of data while controlling costs,” he says.
Question 3: What is the total cost of ownership?
The third question to put to a potential supplier concerns total cost of ownership (TCO), according to Forrester. It is important to make an upfront investment that delivers over time.
“The cost savings from commodity hardware no longer mask the TCO, especially as the volume and complexity of applications, data, and the hardware on which it runs are rapidly increasing,” says the Forrester report.
But it points out that TCO must consider more than just the hardware, which might have worked when business requirements were fairly stable and the price of x86 Linux systems was relatively low.
With real-time analytics requirements, proliferating mobile devices and rapid-fire business demands, it recommends: “Focus less on a two-to-three-year hardware purchase cycle, and focus more on agility, flexibility, and change management.”
Longbottom agrees that the question should not just focus on TCO based on the price of hardware and operating systems, as it can be misleading. He says an organization should ask suppliers: What are the lifetime costs and value of the platforms available? He stresses that value always has to outweigh cost.
“For example, if Platform A has a TCO of £10m and Platform B has a TCO of £20m, which one should I go for? Platform A, of course. However, Platform A gives a business value of £8m, whereas Platform B gives a business value of £200m – that changes the decision quite a bit.”