Observability Costs

A Guide To Observability Costs

In this guide we will examine the pricing models for each of the observability vendors in our sample. Before we start, we are going to cover some very basic aspects of observability system costs. Our very strong recommendation would be that you should not think about the system comparison phase of your procurement operation without having a clear understanding of some very basic measures relating to costs.

Logging Volumes

On the face of it, logging seems like a simple, almost trivial concern. Whilst generating logs maybe simple, ingesting, indexing, storing, querying and retrieving log records with arbitrary structures and multiple formats at large scale is a very considerable technical challenge. It wil be important to have an awareness of two metrics regarding your logging volumes:

The monthly volume of your logs in GB
The total number of log events emitted by your systems

You do not have to know these values with 100% accuracy but you do need a decent approximation.

Metrics

Metrics can be one of the most unpredictable AND expensive elements of your observability costs. A clear understanding of the following concepts is probably essential:

Cardinality
Metrics Series
Points Per minute

Without an understanding of these concepts it can be difficult to make a well-informed costs estimate or be able to compare costs across systems in a meaningful way. There can be a number of difficulties in measuring the volumes of metrics being generated across all of your systems. Partly, the raw data is held in backend systems which may not be easy accessible or queryable. Equally, even if you know the total number of records it is not easy converting that into GB. There will inevitably be an element of guesstimation.

Metrics In The Cost Estimator

When you enter a metrics value in the Core Options section of the Cost Estimator you can enter your data as either:

metrics points per month (in millions)
active data series (in thousands)

You just need to choose the appropriate value from the dropdown list. We will then do the relevant conversions.

Retention Periods

Generally, it is not easy to make cross-product comparisons because policies on retention periods vary so drastically across vendors. Some vendors have fixed periods, others are variable and some offer more granularity than others. Datadog retention can be as low as 15 days whereas Observe set a default of 13 months across logs, metrics and traces. Naturally, if you have specific requirements which exceed the defaults you can normally negotiate a custom arrangement with a vendor. Obviously those rates will vary from customer to customer so we can't include them in our models.

Billing Periods

Most vendors will quote costs as a monthly figure but will assume that have agreed an annual contract. There can be significant cost differentials in paying monthly vs paying annually.

Versions

Most systems come in a number of different versions and pricing tiers. Many will have free versions as well as enterprise versions, for which no standard pricing is published. In this guide, we have adopted the policy of using the middle tier versions of systems and we will specify this when displaying the system name e.g. SigNoz Teams, Honeycomb Pro. These distinctions can be important when you are checking features on the vendors web site.

Complexity and Simplicity

As you use the cost estimator tool and read through our documentation you will notice massive differences in the complexity of the respective pricing models. There can be a tendency to be cynical about the more complex pricing models. It is often argued that the mobile phone industry deliberately uses complex tariffs to confuse users and extract more profit from them. Whilst, for the purposes of this exercise, we are very grateful to those vendors with simple pricing models, there is also another side to the story.

Whilst Observe charge by the terabyte, others may charge on a very much more granular basis. The rationale for this can be the the vendor is aiming to offer more accurate and customer-centric pricing. The principle behind this is rather than averaging costs out over the customer base, each customer only pays for exactly what they use. If you don't need to keep your logs for 12 months then just pay for exactly how many months you want. If you are not running large queries over your traces then you will pay les than a customer who is. That, in any case, is one argument in defence of the more commodified and differentiated approach of some vendors. Whether it really does offer you value is something that you can hopefully decide for yourself.

Comparability

It is worth re-iterating that price is not the only criteria for evaluating a product. Just because one product is more expensive than another does not necessarily mean that it is over-priced. Some products will be more scalable, have more features, better UI, reliability etc. Staying at a five star hotel will be more expensive than staying in a caravan. Neither is better than the other inherently - the question of value depends on the needs of the customer.