base14.io

Fast, comprehensive and secure - roadmap for observability solutions in 2025

What should your observability suite do for you in 2025 ? I guess it's the right time to jot down my expectations from an Observability Suite in 2025. After all, a month is good enough for you to fix the gaps. In 23-24, observability tools grew leaps and bounds, and people finally started looking at observability as something more than just instrumentation. And OpenTelemetry already looks like a very strong alternative to private tools. We've all gained a lot from this change. 2025 should bring in more beneficial changes. Here's what I think your observability suite should look/feel/behave like. 1. Your observability suite should allow you to see all production signals - logs, traces and metrics, in one place. This is because working across multiple tools leads to longer resolution times (MTTD), correlating is easier if all signals are visible on one plane. 2. Your observability suite should monitor all sources of change to the production environment. If you have to find out about a production deployment by calling someone when an alert pops up, if you have to call a vendor to find out if they have an outage going on, that just adds to MTTD as well as human dependencies. At a minimum, your tools should track all infra changes, deployments and vendors' uptime statuses, along with application signals. 3. Your observability suite should store every signal as a wide event - basically record all relevant metadata along with the value of said signal. If you have to go outside of the tool to correlate why an issue on only one OS version or only one node is dropping your time series (orders/transactions/views/etc) you are not on the right platform. *Also called high cardinality*. Additionally you should be able to add an arbitrary number of attributes as metadata for different categories or instances of signals. 4. Your observability tools should be fast, with sub-100ms response times for most dashboards and searches. If dashboards take a minute to load, or alerts take longer to trigger, engineers are already wasting precious time. 5. You should be paying for your observability tool based on consumption. Pricing per headcount or number of hosts etc push you to adopt suboptimal strategies, like giving only half of your staff access, or installing agents in 1/3rd of your machines. You should be able to track data storage expenses in real time.  ....... more here -> https://lnkd.in/gfXV2nUs