PromLTS is a SaaS providing Long Term Storage for the Prometheus open source monitoring system.
Specifically, PromLTS solves the problem of reliably storing a large amount of time series data generated by Prometheus.
The short answer: Adding small remote_read and remote_write sections to your prometheus.yml configuration file.
The short answer: please email email@example.com and we'll get back to you quickly.
The longer answer: There's some more detail on our Getting Help page about how to contact us for different issues, business hours, billing questions, etc.
Our backend stores three redundant copies of each datapoint in our distributed, fault-tolerant, eventually-consistent database. In addition, our backend design is heavily focused on eliminating single points of failure. We're happy to share deep architectural details with our larger customers - please get in touch if that's a thing you'd like to learn more about.
In CAP theorem terms, we provide availability and partition-tolerance.
We've been doing this for a long, long time - longer than Prometheus has been around! PromLTS is run by a company called Metricfire, which has been building and maintaining Hosted Graphite since 2011. We handle billions of datapoints per day from thousands of engineers around the world, and, if we can suppress our naturally Irish urge to be self-deprecating for just a moment, we think we're pretty damn good at managing a huge amount of time-series data on behalf of all our customers.
A lot of big brands trust us with their critical monitoring data, and we hope that you will too. If you have questions about any aspect of our service, please get in touch. We'll give you deeply technical answers about our stack, our security posture, our processes, etc. If you want to set up a call with a founder or with some of our biggest customers to get the real story, send us an email (firstname.lastname@example.org) and let's talk.
Yes. Prometheus, by design, does not downsample any data, which makes it a challenge to store the full 'raw' resolution because it's a huge amount of data to store and process for display when you need to query it back. Past a certain point, downsampling of some kind is always required to keep data volume and query responsiveness within sensible ranges.
Using your local Prometheus in combination with PromLTS gives you the best of both worlds - high precision "raw" data from your local Prometheus for queries on recent data, and fast, accurately downsampled data for queries on data up to two years in the past.
PromLTS supports two downsampled resolutions:
The short answer: 2 years.
The longer answer: Data retention depends on the resolution the data is stored at. It's much less intensive to store and maintain finer-resolution data for a long time, so here's what we store:
By default, your downsampled timeseries data is stored in a large multitenant distributed database.
If data isolation is a serious concern for you, we can offer full hardware-level isolation at the storage layer, or at the storage and querying layers, depending on your needs. Talk to us: email@example.com
We have a security FAQ that we're happy to share with you - please get in touch.
All our servers are in Germany - near Frankfurt, to be more specific. If you have a legal or regulatory reason for needing to store your data somewhere else, please talk to us by emailing firstname.lastname@example.org.