Why Unified Observability Matters for High Performance FinServ Networks (Kashif Hafeez)

Financial Services companies rely on their networks to provide market data to their customers without gaps and facilitate financial trades at high speeds. These networks require extremely low latency – far lower than most enterprises – and must be kept secure.
This places great pressure on the IT teams supporting these networks to maintain these low latencies. At the same time, trends like digital transformation, automation of processes like credit applications, the shift to mobile from desktop, modernizing applications
and personalizing services has made it much harder to maintain this level of performance and security. This has created several unique pain points for financial services IT teams. Let’s explain why these have arisen, walk through some of the most common ones,
and discuss how a strong network observability strategy can help alleviate them.

Changes in Process, Changes in Networks

The rapid growth of high frequency trading in financial markets around the globe is placing new demands on the IT operations teams to ensure high performance, extreme accuracy and robust security. At the same time, more customers are using financial services
like trading via mobile devices, making traffic patterns more unpredictable and harder for IT to manage. It’s inefficient and expensive to centralize apps in a single data center but moving them to the cloud means that IT cannot easily capture the traffic
to and from those applications. If an application performs badly for a user, IT won’t know about it until the user complains, and then will struggle to troubleshoot it without packet data. IT is now walking a tightrope, balancing the increasing expectations
placed on them with the growing complexity of their networks.

Network Performance

Time is money when it comes to financial trading. Fractions of a second and individual network packets matter. High-risk trades rely on fast trade execution and timely market data. Investors and investment managers seek to maximize gains, even from small
price movements. IT must be able to measure network latency to an extremely granular level – down to the millisecond – to identify when it spikes above allowable limits and find and fix the problem. IT must make sure their monitoring hardware (like packet
brokers) are capable of processes packets at 40Gbps or greater speeds (this is not possible without special hardware assistance), and they have the ability to measure, timestamp and analyze latency and jitter at millisecond granularity.

User Experiences

Performance of features matters in a mobile-first world – poor performance means upset customers. But as some service components move to the cloud (or perhaps to multi-cloud environments), the latency of the overall service increases significantly compared
to everything being in the data center (where latency between network components is minimal). The Application and DevOps teams will need to tune new mobile services to account for this and make sure customer gets the same performance, despite all the changes
on the back end. But to do that, they will need a network observability solution in place to measure latency between each service component. It is always a best practice if they test this in advance of the feature going live, and then continue to monitor for
issues once it is in production. The stakes are high – even small amounts of latency are enough to lose customers.

Experience Monitoring

Network Operations (NetOps) teams must monitor KPI metrics (like steady-state traffic, roundtrip timing, bursts/microbursts, latency and jitter) to effectively balance performance, cost and maximize the ROI of the infrastructure. They need to be able to
measure the overall health of their network, track low latency performance metrics across multiple clouds and on-premises, and maintain a holistic view of all operations and execution across the trading cycle, from order to execution. This approach to Network
Operations functions, that focuses on getting a broad view overall network health is often called “network observability.” It allows IT to know when they might need to rebalance loads and when they may need to add new WAN links. Many network monitoring tools
and strategies from the pre-cloud era focus on measuring the performance of individual devices or links and can’t provide this overall view.  

Cybersecurity

Financial trading services are attractive targets for cybercriminals and need to maintain strong network security. While this task falls to the security team, they’ll need the NetOps team to provide them with access to packet data from all parts of the network
to pass through the firewall or feed Network Detection and Response tools. These tools look for signatures of malwares or suspicious network behavior. If IT has detailed visibility into all parts of the network, they can often spot signs of an attack (such
as unusually high traffic on certain ports often used for malware) and flag the issue to the security team for further investigation. 

Troubleshooting all these issues (maintaining security, tracking user experience, fixing issues before they affect those users, and maintaining ultra-low latency) requires the network to be observable to IT – they must have access to full networking data
(including both packets and flow data) from all parts of the network, including cloud and SaaS applications. Above all else, they need this capability – it provides them with the data they need to identify and fix all these issues.

Many IT vendors are now offering more predictive capabilities that surface issues (either through dashboards or automated with big data/AI/ML analytics) proactively, and suggest solutions, which can help relieve the burdens of IT professionals. This new
approach, focused on understanding the network holistically, is called “Network Observability” and it has great potential to help financial services companies meet the punishing demands of their network and keep the services that depend on it running without
issue.