Capacity management “key to operational efficiency”

Companies in the financial markets could save significant operational costs and reduce outages through better management of their IT capacity, an expert has said.

According to Guy Warren, CEO of the ITRS Group, companies across the financial markets are taking on more IT capacity than they need or can cope with resulting in lower margins and the risk of increased downtime.

“There has been in a huge investment in IT over the past decade, but many firms have bought much more capacity than they need.

“We have seen companies in the market running over 50,000 servers but they often have no real time insight into the cold spots and the under used capacity,” he said. “Even if there is only 5% of capacity that is not needed, addressing that can save a company a lot of money.”

Warren said that his clients are increasingly turning to cloud-based servers to optimise costs and increase flexibility.

“Some firms are choosing the hybrid model with some of their capacity held in the cloud and others are going across multiple cloud-based providers such as AWS or Microsoft but as they use a hybrid model or multiple providers they don’t have a total view across all capacity,” he said.

Warren called for a fundamental rethink of capacity planning among financial services companies. Currently firms tend to wait until an application such as CPU, memory, storage or network bandwidth is under high load before looking to increase resources based on which are their limit.

However, this just moves the bottleneck elsewhere in the system according to Amdahl’s Law, which is used to calculate potential increases in speed and capacity across a computer network.

“In order to model complex applications, modelling techniques need to understand Amdahl’s law and have the data to understand how each application behaves in the resources it uses,” said Warren. 

“Without this level of insight, you are in the world of ‘try it and see’, which is both wasteful and potentially disastrous if a peak happens and the resources are not the right ones to achieve the peak load.”

ITRS Group recently acquired the capacity planning business Sumerian, which uses machine learning to optimise allocation of IT resources. The business is being integrated into ITRS’s core product offering of real-time monitoring systems.

Warren believes that the application of machine learning to capacity modelling and to identify and fix breaks across IT infrastructure will massively increase efficiency and reduce downtime.

“If a feed dies today, it can take 15 minutes for a human to identify and fix the issue. That could be done instantly with computer software,” he said.