Will real-time data processing replace batch processing?
At Confluent's user conference, Kafka co-creator Jay Kreps argued that stream processing would eventually supplant traditional methods of batch processing altogether.
Absolutely: Businesses operate in real-time and are looking to move their IT systems to real-time capabilities.
Eventually: Enterprises will adopt technology slowly, so batch processing will be around for several more years.
No way: Stream processing is a niche, and there will always be cases where batch processing is the only option.
AI / DevOps / Operations

Best Practices for Mastering the Incident Life Cycle

Improving efficiency requires a codified process and an emphasis on moving from a reactive to a proactive mindset.
Aug 3rd, 2023 7:22am by
Featued image for: Best Practices for Mastering the Incident Life Cycle
Image from Jaromir Chalabala on Shutterstock.

The complexity of today’s technical ecosystems is unmatched, and the pace is breakneck to deliver revenue-driving innovations. Success is measured in days and weeks, not months or quarters. Yet many organizations have the task of delivering this innovation and pleasing customers with the same or even diminished resources.

In the do-more-with-less era, organizations are looking to AI and automation to improve their efficiency. One area for optimization with the biggest possible gains is incident response. If responders spend more time innovating and less time firefighting, they can deliver value faster. And with fewer interruptions to service and better digital experiences, organizations will see improved customer loyalty.

Achieving this isn’t as easy as dropping prompts into your generative AI tool of choice, however. It requires a codified process and an emphasis on moving from reactive to proactive incident response.

Reactive vs. Proactive 

A reactive approach to incident response is characterized as chaotic, ad-hoc and highly manual. The same problems may occur again and again, and there’s substantial customer impact. This results in exhausted teams that are continuously firefighting rather than resting or innovating. It also means teams can be interrupted outside of their normal working hours, something that affects 54% of responders.

Moving toward a proactive approach is about preventing humans from being the first line of defense and mitigating customer impact. Automated incident response introduces methods to leverage machines to shoulder some of the burden and help humans balance critical workloads. A proactive approach also requires learning and iterating to improve system reliability over time.

A key part of this transition is understanding the incident response process and scaling it across teams.

Own the Process

Before outlining the stages of the incident life cycle, there’s some important groundwork organizations need to complete first. Collectively, all teams need to align on the answer to three important questions:

  • What is an incident? Distinguish between day-to-day maintenance issues and incidents that affect customers by giving teams a framework with set criteria for each categorization.
  • Who does what during an incident? Define clear roles and responsibilities for the people involved in the response. Common roles include incident commander, deputy, scribe, internal liaison, customer liaison and subject-matter expert.
  • What tools do we use? Understand which tools your team uses during incident response. Do you look at certain monitoring tools? What do you use to communicate with stakeholders and customers? What do you use for collaboration? Do you have an IT service management tool that needs updating? Building a toolkit can help teams work faster by ensuring all members are on the same page.

Next, organizations must understand what happens during each phase of the incident and look for opportunities where automation and AI can help ease the burden on responders.

1. Detect

Detection can take many forms. Ideally, your monitoring tools identify anomalous behavior and transmit the data to your first-line responder. But you can detect incidents in other ways. Sometimes customers notify support agents of an issue, or your technology teams might detect an incident that wasn’t picked up by your monitoring tools. It’s important that these incidents are immediately delegated to the right people. Organizations can consider building an automation that routes the alert based on set criteria derived from event data for faster MTTA (mean time to acknowledge) and fewer escalations or reassignments.

2. Prevent

When an incident occurs, there’s often more data than responders know what to do with. Alerts come in rapidly, creating an alert storm. During this phase, it’s essential to avoid excessive noise so that the responder can concentrate on the important task. This could involve using machine learning to silence transient alerts that typically auto-resolve themselves. It could also be a perfect stage to introduce auto-remediation for well-understood and well-documented incidents.

3. Mobilize

Once it’s clear that a human needs to act, it’s time to mobilize the response. You need the right people working on the incident and the right processes in place to help them work efficiently. Incident workflows can help by kicking off processes, adding responders and spinning up collaboration tools. This automation reduces toil for a responder and improves MTTR (mean time to resolution).

4. Diagnose

In the diagnosis phase, the right people look at the problem and follow established processes. For this to happen, you need information quickly. However, it can be difficult to know where to begin and toilsome to run routine diagnostics manually. But tools exist to relieve this burden. For example, with AI, teams can look at past and related incidents and use contextual information to help them resolve incidents faster. And through automated diagnostics, teams can trigger basic actions such as getting current status on memory, CPU health or gathering error logs.

5. Resolve

Resolution is often the longest and most difficult process. Here, teams work to bring the incident to a close. But incident responders need to also communicate constantly with other teams, business stakeholders and even customers. It’s important that the response team sets clear expectations with stakeholders, and that these updates are simple. After all, if a responder is constantly fielding questions, they’re not able to resolve the incident as quickly. Crafting status-update templates that are reusable can save responders precious time. And using generative AI to create these templates can speed up the process further, preserving responders’ cognitive capacity for fixing the problem.

6. Learn

Even once resolved, an incident isn’t “over.” Teams need to take time to learn and reflect. With the help of a comprehensive analytics tool, teams can run customized reports on areas such as responder data or noisy services. Such reports can help teams better understand processes to identify areas for improvement. By incorporating learnings, teams can respond better to future incidents and implement upgrades to processes that will benefit the entire organization.

Putting It into Action

With the right investments at each stage of the incident response life cycle, organizations can use AI and automation to boost their efficiency. But the road to proactive incident response is a long one. Change doesn’t happen overnight. Organizations should focus on incremental changes that improve the reliability of the system and processes and that give back time to responders. Doing more with less is a superpower, and the companies that can harness AI and automation to help them codify, standardize and improve their incident response will be the ones who win their market and impress their customers.

Group Created with Sketch.
TNS owner Insight Partners is an investor in: Resolve, Pragma.
THE NEW STACK UPDATE A newsletter digest of the week’s most important stories & analyses.