Influx vs Prometheus vs Timescale 

Yeah that’s precisely why I wanted to try it, it feels like the path of least resistance. Although, if I have to pick one as the one I feel has the most potential to grow and dominate, I think it’ll be influx, it’s just my hunch though.

Thank you for taking the time and leaving a compliment! I’ll make more videos when I’m back from my trip

TimescaleDB is the one I have little experience with, though it’s the one I wanted to use, it’s a pity that managed Postgres in AWS doesn’t support it and my company is too large to move all the parts to sign a new contract.

Having said that, my experience with influx so far is great. Nice performance, flux is easy once you get it. Curious to see what influx IOx can do.

Les debo ese video hace mucho. Estoy cerrando una release y me pongo a grabar

Hey thanks for the comment! I haven't had the chance yet. My guess is that with SQL query language it's going to be more accessible for developers. One of the biggest barriers I have in my company with influx or prometheus is explaining how they must create their schema to avoid cardinality in tags/labels and the query language (flux/promql)

I like influx a lot so I’m biased. In this case I think either one would work fine, I personally would go for influx. Even Prometheus gauges here would be fine since you are ok with just periodic data.

Hey thanks! appreciate the compliment. I’m using Final Cut Pro, but it’s overkill for the kind of editing that I do. I should have gone for a simpler tool. I use like 0.5% of the features it has.

@@Gabzim i get you, but i think most tools out there are pretty overkills, although tbh Final Cut Pro is pretty simple and straightforward to use Keep the videos coming!

They Tom thanks for the kind words and I’m glad it’s helping you. I had a very complete series planned for this topic using examples from my real job. Unfortunately I’m a bit of an introvert so it takes a lot of energy for me to record myself and talk to a camera, I’ll try to put a bit more effort in if people find these helpful

@@Gabzim I think the use of Grafana for BI is gaining popularity and observability is the new trend right now. TSDBs can be very confusing and intimidating and this comes from a guy that has been doing rdbms for 20+ years. People really just want a BI/reporting/visualization tool that is dead simple and they can generate pretty graphs/dashboads in minutes and not have to steep learning curve such as power bi/tableau/looker. Looking forward to your next videos.

@Tom N agree, I’ll try to make a couple this week. Thanks for the support.

I’d use influx for that.

Flux also has other features that will be very useful for analyzing that data. Stick around, I’ll make a video on that.

@@Gabzim thanks man! Great video btw 👍🏿

@@Gabzim thank you, waiting for the next

I'd go with QuestDB, which was designed specifically for this use case. It is used now for every kind of time-series data, but fintech/trading/crypto are some of the more popular use cases amongst its users

Ah seems to be something of ios.. Anyways great explanation keep it up

GoodNotes and I just record my screen

That's ok because, when you query the data, the rate operation in prometheus takes this into account. Let's say you have a counter that goes like: 1, 3, 5, 8, 10, 2. That last 2 is "wrong" in the sense that a counter should always increase, so rate will look at the data like: 1, 3, 5, 8, 10, 12 (roughly speaking) and will calculate the rate from there.

Because then each file would create a new metric with a single data point, after a while this would really wreck your performance and you’d need to accommodate for this in your queries. Labels have to have a low cardinality

@@Gabzim labels are used like that in loads of places though, and pretty common right? e.g. node exporter, so it is possible, but like with all prometheus metrics, you just have to be wary of cardinality - though it feels like the same could be said when comparing to some downsides of SQL based systems - just feels like in this case its maybe a bad example to give of something you can't do in prometheus, when you can? Side note - really like your drawings, it fits in really well with the video as you explain stuff.

Of the three, influx is the one I’d personally recommend. You can’t push data to Prometheus so you’ll lose resolution.

In addition, flux has multiple functions that are useful for processing moving averages, predictions etc

if you mean via the push gateway i’d say “kind of”. you can push it to the push gateway but then that gets scraped by prometheus. if you mean you can push granular data in between scraping intervals i haven’t seen that.

Sorry, I switched jobs and started a side project and I couldn’t keep up with everything, I’m hoping to take some time off soon and make videos

Hey brother, sorry about that. My life changed radically and I’ve been very bad at heading a company plus being a father plus RU-vid. I really should organize things and get this done.

Really appreciate the kind words btw

Valid point, sorry about that.