Translate Datadog Metrics into OTLP | Hacker News Reader

Translate Datadog Metrics into OTLP

grafana.com

74 pointsburo910 comments

This is really exciting since the Otel collector already supports Datadog tracing, adding metrics will cover the big two signals that make it hard for most teams to shift from DD's walled garden. I suspect this will really ease the migration path from closed standards to open for a whole swath of companies now.

Hopefully the PRs get upstreamed quickly without any opposition from Datadog unlike last time.

std disclaimer: I'm building oss obs for hyperdx, so I have a clear horse in the race for companies adopting open standards, just like grafana.

Wait, the Otel collector supports Datadog tracing? Do you have a link to docs? I tried looking into this a few months ago but couldn't find anything.

Yes, in the collector-contrib repository you already had support for traces https://github.com/open-telemetry/opentelemetry-collector-co...

I'd be curious to know how many organizations once they get on DataDog get off of it, despite it probably easily costing $40k/mo between log ingestion/indexing, span/trace/APM storage, monitors, etc. for smaller tech companies.

Yes, you can do it all yourself open-source, invest engineer time to keep it all running. Pay for cloud infrastructure, run DataDog-like/clone UIs...

There's just something about paying top dollar to get the "crem de la crem". That isn't to say that there's just a certain level of polish missing from all open-source alternatives (some of them probably even strive to be drop-in replacements/turnkey)

but DataDog is a $40b company for a reason I guess?

That 1-5% of polish missing from opensource is worth $40b I guess? Is that a good number to represent "what is missing in 'free' solutions versus DataDog"? 5%? Might be underselling it.

I think the big reason they're so huge is they make it easy to onboard alllll your data, and incentivise you financially to stick with all their custom tooling and instrumentation. So the "quick wins" have a long tail of financial scaling. As your own scale grows, your DD costs grow. And so you do a few things that may entrench you further to drop your costs and the sunk cost fallacy helps corner you. You need your dashboards and alerts now that you're bigger, so you aren't thrilled about the prospect of rebuilding from scratch.

Anyway, I know this because I've been there at a former employer. We started off (and mostly stayed) otel at the base layers, but more and more had to move to proprietary DD collectors/agents for off-the-shelf system metrics (db, cloud, VMs, k8s, etc), because otherwise they were custom metrics and very very expensive.

So they're very very sticky. And the market loves a sticky + expensive vendor.

> Yes, you can do it all yourself open-source, invest engineer time to keep it all running. Pay for cloud infrastructure, run DataDog-like/clone UIs

This stuff is table stakes for any competent ops person and we can spin this whole stack up in our sleep. You do it at every job. It's cheap to run and maintain as well. This isn't to say that Datadog can't be worth it but $40k/mo will get you the whole stack plus 35 hrs/wk each of a senior and junior SRE that can work on other projects.

If you're swimming in cash and want the best at any price then hell yeah do it, DD is unbelievably polished. But if you're at all cost sensitive it's a bit of a catch-22 because if you have enough infrastructure to warrant DD then you likely have an ops team that can do it themselves.

a lot of the open source stuff has click to deploy setups in AWS & GCP too

This has been a constant source of friction between build vs buy, esp. when talking about day 1 operations, which is why at OpsVerse (https://opsverse.io/), built out ObserveNow, a fully managed, OSS based observability platform built for enterprise scale. So you don't have to tie yourself into a walled garden, can run pretty much any exporter to bring in data to the system (or take out), and always have access to the tooling that powers your platform without needing to babysit the observability platform itself as you grow.

And since it's purely ingestion based pricing, in general it works out a LOT cheaper as well.

explicit disclaimer: I work at OpsVerse

Grafana has hosted cloud offerings for metrics/logs/traces and some other stuff that come in waaaaay under DataDog costs.

The super-integrated UI polish type work isn't up to DataDog's level but considering it's a way cheaper option to still not have to run it yourself, it's a good option.

All of these threads turn into how expensive Datadog is and discussions about open source alternatives.

My 2c - observability is the easiest thing to rent. There are zero reasons to build it out and have a team maintain it. I cannot imagine a management chain wanting to invest in a greenfield unpredictable and guaranteed unreliable 0lly service. Datadog's great super stickiness comes from their polish, their availability, and turn key instrumentation that is hard to build from scratch.

For any open source alternative to compete, they need to be at least similarly turnkey.

The only real threat to Datadog are other closed-source competitors who are

1. Willing to invest to be as reliable and turnkey

2. Willing to accept a lower price. Literally price it $1/host less than Datadog and a lot of companies will start the migration.

Reply on news.ycombinator.com