With Peak Shopping Periods Around the Corner, What’s Your Observability Strategy?

Summer is winding down, which means online retailers are once again taking steps to ensure their website operations are sufficiently bulked up to handle the approaching traffic onslaught. They're focusing on scalability, elasticity, and increasingly observability — which is emerging as a hot buzzword lately.

Observability is growing in importance given today’s sprawling application and infrastructure landscapes. While a typical enterprise may have dozens of traditional monitoring tools for various apps and systems, this can result in data silos that lessen the effectiveness of an overall monitoring strategy. Observability, on the other hand, is achieved when an organization has pooled all the data (logs) it needs to proactively detect and resolve emerging issues. In essence, it allows the online retailer to understand the overall health of applications and infrastructure at a glance and unearth critical correlations enabling the web operation to run like a well-oiled machine.

For example, an organization can group all logs related to the cart checkout process, across various apps and systems, to identify a shared slowdown in response times and increase in cart abandonments. This would point ops teams in the right direction in their hunt for a root cause, such as a bottleneck in the checkout process. The traditional approach to observability has been to index and crunch huge amounts of log data in a central repository. The more data, the better, as greater data richness leads to more valuable contextual findings. However, this approach isn't as simple as it seems and presents several challenges that organizations must consider as they evaluate their observability strategy.

How Much Observability Do We Actually Need?

Can you get away with neglecting certain data sets? Perhaps. But every time you do that, you’re taking a chance and opening yourself up to a blind spot. This holds even more true during peak sales periods when your apps and systems are generating data at a rampant pace.

In addition, choosing to keep certain data sets while neglecting others is just not suitable to the nature of many e-commerce applications, which are increasingly multistep (all working independently, but together) or even containerized (and subject to constant shifts). All this can make such decisions very haphazard and risky. The natural answer is to keep all of your data, but you must consider the storage cost implications.

Do We Have the Manpower to Appropriately Analyze All This Data and Extract Value From it?

Observability tools do a great job providing access and drill-down capabilities into huge amounts of useful log data. However, this data can quickly dissolve into an overwhelming and confusing fog, as more data is emitted across various disparate applications, components and systems.

Where's Our Data Being Generated?

Edge computing offers many benefits to online retailers, including delivering more cost-effective and real-time insights on customer behavior, which can be used to improve marketing and drive conversions. In e-commerce, speed matters. Statistics show that even a .1 second improvement in speed improves customer engagement at every level, including sizable increases in conversions and order values.

Speed also matters when it comes to managing logs in order to optimize site availability and performance. For an online retailer leveraging edge infrastructures — e.g., regional cloud data centers or CDNs — the process of centralizing and indexing data for analysis creates latency and is just not as agile and expeditious as a mission-critical environment requires.

The reality is that when you neglect a data set, any data set, you’re stepping into dicey territory. New approaches are needed to help online retailers avoid painful (and potentially costly) decisions, empowering them to fully leverage their rich data sets while keeping a watchful eye on costs, manpower and speed.

Ozan Unlu is the CEO of Edge Delta, an observability automation platform that shows a complete picture of application and service behavior.