The question of performance has become a key concern for IT teams. First of all, there is the obvious matter of the raw performance of the information system, which often determines the level of success of projects undertaken by the IT department.
This is even more true within the current context of digital transformation, “which now involves more than 80% of European companies,” explains Ludovic Levé, managing director for France at Lenovo Data Center Group. Previously, technology was only a work tool. Nowadays, it affects strategies, business models, innovation and overall business performance.
Right now, information systems are “a complex and interconnected network of applications, interfaces and databases where numerous elements have to communicate with external systems and APIs,” points out Matteo Pacca, senior partner at the Paris office of consultancy firm, McKinsey & Company. Companies need to find ways to make these patchwork setups perform while tightly controlling costs and maintaining legacy compatibility.
Uniform use is not an option
Fabrice Benaut, former CIO of market research company GFK, agrees. “I worked at the company for 25 years; towards the end, I was managing 60 countries, which had integration problems within the group, where they were making purchases and major investments, without being sure that the relevant professionals were involved.”
Benaut believes that IT managers do not have the option of uniformly implementing systems when those in use are very different from each other. “So, what can we do? When faced with this issue, you have to see more opportunities than constraints. This is an opportunity to develop the information system. However, it is also important to employ common sense and take use cases into account. There’s no point in moving quickly if you’re headed in the wrong direction,” he adds.
This is something that Christophe Puzenat understands very well. As the head of the technical architecture staff at Solvay, a large chemical group, he was faced with an aging RAID 5 infrastructure that was no longer fit for purpose. Even worse, the EMC solutions he was using were reaching the end of their support lifecycle.
He decided to switch to a completely new infrastructure based on virtual environments, using Veeam and VMware. Once underway, the performance tests gave results that were a hundred times better than the former infrastructure. For example, the maximum delay observed went from around 300ms to 6ms. This change also opened the door for more flexibility in infrastructure management. “Users can now ask me for a few TB for the next day,” he exclaims.
Hyperconvergence to support performance
But not everyone can rebuild an architecture from the ground up. The use of hyperconverged solutions also appears to have found a lot of popularity as a solution to these problems. Systems capable of amassing network, storage and calculation-power problems at a single point are being pushed more and more by suppliers. Such is the case with Lenovo, which has entered into a partnership with Nutanix to supply their hyperconverged HX solutions.
At Criteo, a French company specialising in web marketing, performance isn’t a question, but rather a requirement. “Criteo predicts the future within a very specific domain. Our success comes from our knowledge of purchase probabilities, which we calculate based on previous behaviour of internet users,” explains Nicolas Helleringer, Criteo engineering director.
He outlines his problems: “In the case of bidding for advertising space on a media site, whenever a user is about to connect, the media will query a certain number of advertising transmission networks, which includes us. We have a maximum of 100ms. to reply and choose which ad to display to the user, by specifying how much we are willing to invest to have our ad displayed.” To meet this requirement, the company processes 150 billion requests every day with a response speed of around 4.5ms. This puts huge pressure on the infrastructure.
Performance at the heart of the user experience
Performance also plays a major role in user experience, which has become a crucial issue for businesses. A very good example of this is the online video platform Dailymotion, which has to deal with direct competition from the YouTube behemoth. “We can’t allow ourselves to have videos that are constantly stuttering,” explains Magnus Näslund, company program director. He has chosen to set up his content delivery network infrastructures to ensure optimum service quality on every device throughout the world.
“We can now determine the best route, and switch between the different paths available in real time so as to offer the best possible experience,” says Näslund, “we also have people we call on and a lot of automated supervision using a number of KPIs that monitor use of the platform and any possible technical issues in real time.”
A measure that is just as important
It isn’t just performance that matters; measuring it is just as important. “Whether you’re in Paris, Singapore, Hong Kong or Pretoria, there are no variations, it’s the same experience everywhere. Tests are used to measure this availability, sometimes with outside partners, but also with human testers, or using automated procedures,” explains Näslund. The challenge is to constantly keep track of this performance using real-time monitoring and supervisory tools.
The use of this type of solution has become indispensable for monitoring performance. Software-defined networking tools already provide a good visualisation of infrastructure status, and are currently taken care of directly by manufacturers, such as Lenovo. However, the use of monitoring tools, such as Splunk, is too useful an asset to ignore, and Lenovo has developed its API XClarity to make it easier to connect its infrastructure solutions to Splunk.
The solutions are out there
At a time when digital technology is the key to business success, the performance of IT infrastructure is essential, and so is monitoring it. “It is our job to supply high-performing infrastructures that meet the needs of our clients. As for monitoring, we’re engaging in partnerships, such as the one we have with Splunk, to give clients the ability to measure and monitor this performance,” says Data Center Group’s Ludovic Levé.