Hello Brian, Thank you, this makes a lot of sense. Lifesaver. I think my confusion came from a combination of:
This statement: So PromParser.Metric for example looks to be the length of the full > timeseries name This statement: and on top of all that the time series name twice This image: [image: Screenshot from 2020-03-12 18-35-44.png] <about:invalid#zClosurez> But reading what you've stated make this all much more sensible and logical. Thank you again! Regards, Aaron Layfield On Thursday, 12 March 2020 18:05:10 UTC+9, Brian Brazil wrote: > > On Thu, 12 Mar 2020 at 08:52, Aaron Layfield <[email protected] > <javascript:>> wrote: > >> I've gone over Brians post here that provides a handy calculator that >> makes it easy to calculate rough estimates on consumption: >> >> >> https://www.robustperception.io/how-much-ram-does-prometheus-2-x-need-for-cardinality-and-ingestion >> >> However, I am trying to break this formula down into something I can use >> outside of the page and I'm struggling a bit. Based on the calculator in >> the post itself. >> >> Based on the article: >> >> This works out then as about 732B per series, another 32B per label pair, >>> 120B per unique label value and on top of all that the time series name >>> twice. Last, but not least, all of that must be doubled given how Go >>> garbage collection works. >> >> >> According to the image in the post, time series name (PromParser.Metric) >> is 320B. >> >> So I'm interpreting this as, based on the default 1,000,000 time series >> in the calculator (*Ignoring Unique Labels as those are negligible)*: >> >> Avg. Timeseries Size: 732B >> # of Series: 1000000 >> Label Size: 20B (5 Labels) >> Timeseries Name: 320B >> >> ((732+320+320+100)×2) × 1000000 = 2944000000 >> > > I think you're double counting there, and I'm not sure how you're getting > 320B for the time series name. The exact formula used (field numbers are > off by one) is: > > CEIL((fieldname2 * (732 + fieldname3*32 + fieldname3 * fieldname5 * 2) + > 120 * fieldname4) * 2 / 1024 / 1024) > > So (1000000 * (732 + 5*32 + 5*20*2) + 120 * 0) * 2 = 2184000000 > > > Keep in mind this is for 2.9.2, the numbers will have changed a bit since > then. > > Brian > > >> >> This is way above what's suggested (2GB per 1,000,000 records) and I'm >> certain I'm missing something. >> >> Can someone break this down for me so it makes more sense? >> >> Kind regards, >> Aaron >> >> -- >> You received this message because you are subscribed to the Google Groups >> "Prometheus Users" group. >> To unsubscribe from this group and stop receiving emails from it, send an >> email to [email protected] <javascript:>. >> To view this discussion on the web visit >> https://groups.google.com/d/msgid/prometheus-users/56bccff9-4986-4f7c-ab8c-df89f372436a%40googlegroups.com >> >> <https://groups.google.com/d/msgid/prometheus-users/56bccff9-4986-4f7c-ab8c-df89f372436a%40googlegroups.com?utm_medium=email&utm_source=footer> >> . >> > > > -- > Brian Brazil > www.robustperception.io > -- You received this message because you are subscribed to the Google Groups "Prometheus Users" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion on the web visit https://groups.google.com/d/msgid/prometheus-users/570d8b63-afdb-408e-9437-e5f1b53a8244%40googlegroups.com.
