A colleague of mine recently embarked on a journey to explore the capabilities of a well-known legacy observability platform within his Kubernetes environment. He dedicated a week to familiarize himself with the platform, primarily testing out the different features for traces, logs, and infrastructure monitoring. However, his focus shifted when a critical feature needed an early release, diverting his attention away from the observability tool. Unfortunately, without any prior notification or warning, there was no rate limitation to the platform logs collection mechanism. One line of YAML configuration file meant all logs were collected, ingested and stored — with no mention of the projected cost.
Fast forward to the following week, a member of the billing department barged into his office, demanding an explanation for an astronomical observability bill totaling $33,000 for a single month, a staggering contrast to the anticipated $1,700.
This series of events left my work buddy struggling with the size of his mistake, and me questioning whether it really was entirely his fault.
The Complex Landscape of Observability Pricing
Navigating observability pricing models can be compared to solving a perplexing puzzle which includes financial variables and contractual intricacies. Predicting all potential costs in advance becomes an elusive endeavor, exemplified by a recent eye-popping $65 million observability bill.
Avoiding miscalculations as the one that happened to my friend requires continuous monitoring of the monitoring solution. This practice slows down day-to-day operations and long-term growth efforts.
The Challenge of Affordability in Observability
The escalating costs associated with observability represent a vast challenge which is confronting many organizations currently. Particularly in the age of cloud computing, IT leaders and even top executives have come to realize the imperative of reining in their infrastructure budgets, which often spiral out of control.
The proliferation of microservices and distributed architectures has ushered in a flood of data that demands observability. Traditionally, more data translates into higher expenses, accompanied by substantial resource consumption, leading not only to increased costs but also inefficiencies.
Regrettably, most observability tools employ pricing models that defy prediction. While applications generate large amounts of log data, instead of an advantage, this abundance has become a cause for concern. In response, best practices now advocate monitoring "only what you need" or limiting the retention period for collected data to a minimum. This raises two questions: how can you know in advance what you will need, and will limiting the retention period to a minimum make it impossible to correlate with out-of-range historical data.
Enter eBPF: A Game-Changer
eBPF (extended Berkeley Packet Filter) has recently emerged as a revolutionary technology that has significantly impacted the Linux kernell. eBPF operates at specific hook points within the kernel, extracting data with minimal overhead, safeguarding the application's resources from excessive consumption. It observes every packet entering or exiting the host, mapping them to processes or containers running on the host, thereby offering granular insights into network traffic.
Moreover, eBPF-powered agents operate independently of the primary application being monitored, ensuring minimal impact on microservice resources.
The combination of visibility depth and stability has made eBPF a groundbreaking technology for cybersecurity companies, and is predicted to have the same effect on observability, for exactly the same reasons.
Hassle-Free Observability
Observability should empower engineers, not bury them in a load of unexpected overheads, data volume surges, and huge subscription bills. The goal of observability platforms should be to guarantee complete protection against such surprises, offering immunity against sudden spikes in data volume and shielding engineers from unfortunate encounters with the billing department.
In conclusion, the journey to achieving efficient and cost-effective observability is full of challenges, but with the right tools and strategies, IT and DevOps leaders can help their organizations emerge from financial uncertainty and empower their engineers to become true observability heroes.