techBytes #11: Detective stories for developers: The art of technical analysis

[in-person event]

Hello everyone,

Every engineer has been there. A dashboard suddenly turns red. Something stops working in production. Nothing is obvious, data is incomplete and sometimes the only way to move forward is to dig deeper and follow the clues.

That’s when you have to switch the keyboard to a magnifying glass and put your detective cape on.

We’re excited to invite you to our next techBytes meetup at Ocado Technology. This time we’re focusing on the detective side of software engineering. Across three talks we’ll take a deep dive at real stories from production incidents, performance investigations and everyday engineering decisions. Join us to meet fellow detectives and add new items and strategies to your toolbox.

***Agenda***

18:00 - Grab a drink and take a seat

18:05 - Intro

18:10 - 18:40

Speaker: Maciek Józefczyk, Staff Cloud Engineer at Ocado

Topic: SRE in Action Case Study from Incident Management

Abstract:

Everything looks fine - until your phone wakes you up in the middle of the night. And the data you need to understand why is not there.

Several hours later, after digging through the infrastructure and several plot twists, detective Maciek and the team emerge with a solution.

This presentation shows how incident investigation works in practice. Maciek will walk you through a retrospective analysis of a production issue, focusing on an SRE case study, where the key diagnostic data was not available at the start. The team had to investigate the problem with limited observability and missing logs, while also considering how much they are willing to pay for better proof.

18:40 - 18:45

BREAK

18:45 - 19:15

Speaker: Jacek Kaczmarek, Principal Software Engineer at Ocado

Topic: Speeding up Spring Boot application start by a factor of 2 or 3 at enterprise scale and what we learned

Abstract:

Can you improve the application startup time 2-3 times across 500+ microservices running on dozens of AWS accounts? Absolutely - if you own a centralised platform, a standard technology stack, apply the right tools and data.

In this talk Jacek shares a multi-year (2024-26) investigation into Spring Boot startup performance which led to practical 2 to 3 times improvements in production systems. The work is based on systematic measurements, research methods and iterative analysis, with a strong focus on pattern recognition and identifying both suspected and unsuspected sources of lag. The investigation looks at sources of variance, including hardware differences and autoscaling behaviour, as well as the impact of JVM advancements in recent Java versions from 21 to 25 in enterprise-grade applications.

The session also covers tools and techniques such as JarLauncher, AppCDS, AOT and GraalVM, showing their practical use, limitations and trade-offs beyond typical demo scenarios.

19:15 - 19:20

BREAK

19:20 - 19:50

Speaker: Paweł Stobiecki, Staff Software Engineer at Ocado

Topic: You need far less data than you think

Abstract:

Let’s talk about a queue. But not the AWS or GCP one. The actual queue. Your ultimate nemesis when you and your other half walk towards your favourite restaurant without a booking on a cold November afternoon.

While you try to size it up, the crucial question appears in your head:

Should you join the line, or give up and go somewhere else?

'That's a coin flip', you may say. 'I have no data to decide better!'

And that’s where detective work comes in. Paweł will show you an alternative approach involving three steps:

- realising what you’re really asking about,

- estimates (