Welcome to Composable Observability

We at Mydecisive.ai are so excited to share with the world our vision of composable observability.

First, let’s talk about our definition of observability. Then, let’s get on the same page about the power of composition in observability. Last, we can talk about our technical approach through an example.

<Shocking statement> Observability is not the ability to understand what a system is doing without knowing its implementation details. </Shocking statement>

Let’s keep it simple and define observability via comparison to other approaches:

Definition	Others	Us
Making your things observable	Use vendor-specific agents most of the time / for best experience	Bring your own agent (vendor and/or OTel)
Observing	Fixed data (Metrics, logs, traces) for engineers	Every aspect of systems, people, and software lifecycle for intelligent automation loops
Use cases	Visibility during an incident	Full IT automation and logic loops
Applicability	App instances	Any high-speed high-cardinality data stream
Design	Cloud blackbox SaaS	In your cloud, as a self-managing Kubernetes cluster
Interface	Proprietary queries, dashboards, and manual data scraping	OpenTelemetry, OTLP to our core
AI capabilities	“Smart” dashboarding	AI for intelligent automation of scaling, cost savings, and incident response

The MyDecisive Observability Definition

To observe a component of your stack is to understand its full state so you can take informed automated actions. Full state includes what it is, how it was developed, as well as how it is behaving at any given time, who put it there, when they did so, as well as when it behaved a certain way.

We think that observability is the practice of leveraging the “5 W’s” (WHO / WHAT / WHERE / WHEN / WHY) around your components to take actions every day. We are not another incident response tool, nor are we a dashboarding tool.

If you can work with the state of your systems (the 5 W’s) and act on that state as well as changes in state, then you can truly start to compose intelligent IT automations.

The Power of Composition

Composition means that you can take your data at any point in your observability pipelines and do valuable things with that data, including:

• Changing the data / rewriting it
• Enriching the data by JOINing it with other data
• Routing where it goes
• Storing it in various forms for various use cases, from compliance to dashboarding to AI use cases
• Filtering it to keep just what’s valuable and transmitting it based on your budget

With composition as a platform offering, you don’t get a single capability – you get them all at once. If you want to make sure one of your applications never transmits more than 10GB per day to a particular vendor, and you simultaneously want to scan all data for PII data before transmitting it, and you want to integrate change control processes within your ITSM to CI/CD pipelines, you can do all 3 at once on our platform, even on the same data stream.

Our Favorite Example

Our favorite example of composable observability is what we call “budget-based logging.” Some vendors offer static drop-rules for logs that help you save a small amount of money. Recently, vendors have added AI-based log compression that deduplicates log entries and amplifies what may be critical signals for you. Whether AI or static drop-rules, all these solutions leave much to be desired. Namely, you have too little control over which logs get into your vendor solution, you are paying too much for logs you don’t use, and there is too much maintenance of the solutions required to keep budgets in check.

Budget-based logging means that you first put all your teams and organizations into the MDAI platform. Then you tell us which teams own what services via runtime tags or rules or a mix of both. Last, you tell us the overall enterprise budget for logging and the per-team line item budgets, and we can maximize the number of logs sent by each team without going over budget, both at a team as well as at an enterprise level.

How Is This Better Than Alternatives?

There are three critical benefits to the MyDecisive approach in this example:

1. First, we have all the metadata separated from the filter logic so ownership and organization structure can change without having to update your filters. Budgets can change separately from teams. Basically, everything is declared instead of hard-coded in our system, so it is very easy to change and live with our solution.
2. Second, you may want to deploy different filter logic on a per-service basis. How would you write pipelines that handle multiple filter scenarios where some services send no more logs when they’ve maxed out their budgets, while others must always send WARN and CRIT level logs, and still others do something entirely different? Managing one or two different filter pipelines for a series of 100-1000 services might be ok, but mapping 10 or more filters to 1000s of services and handling change becomes overwhelming. This is a prime example of the power of declarative streaming data management at the heart of composable observability.
3. Third, you can layer budget-based filtering with our data filtration or log tiering solutions so that when we have to constrain an app’s logs due to budget overruns, we can still place the logs somewhere you chose instead of dropping them as other solutions do – Loki would be an open and easy example.

That’s right. Full composability means assembling just the right combination of capabilities to support your need to maintain budget, and yet still have logs available during an incident. And, critically, composable observability means truly tailored solutions to fit every need of every team and service simultaneously under one umbrella solution. After all, many apps will have many requirements, and what better way to solve them than with a configurable solution designed with all the capabilities as plug-ins?

With MyDecisive, you will finally be able to tailor your data to meet your needs and do it at the lowest possible cost.

Welcome to the age of total telemetry control with MyDecisive.ai.