Damn, my measure was wrong. We are producing/consuming average 1 million messages per day (data collected for a week) – average 700 per second. Server never shows more than 5% of CPU, and memory is stable at . No high usage, and server behaves really well (we expect much more on few months from now), as said already in this thread, performance and benchmark is something you need to do with your scenario in mind. Just setup your environment, and run some simulations.
Before putting in production, put everything in a staging and do some load test near to real production as far as you can. You will be safe and confident by then. I like to push staging until systems break (OS, MQ, or my microservices); then I know the limits (or what else to look for). There are lot more components that can fail, not only Artemis... For example, in one of these staging tests, we’ve discovered the major problem was datacenter badnwidth we buy’d was too low. Buying more bandwidth solved the problem. Regards, ER Enviado do Email<https://go.microsoft.com/fwlink/?LinkId=550986> para Windows De: Clebert Suconic<mailto:clebert.suco...@gmail.com> Enviado:quinta-feira, 29 de setembro de 2022 07:10 Para: users@activemq.apache.org<mailto:users@activemq.apache.org> Assunto: Re: Is Artemis Production Ready? > > 5k / second is fairly low IMO. But you have to say how you are producing and consuming ? If you create a producer with an async callback (to answer your client of completions) you can easily obtain 100k / second. (With some avg hardware ) If you have a web service ingesting data and you block the Http Post you will be using more threads and may not scale the ingestion. (Even thought the broker would be ready to process more simultaneous requests ) It’s all up to how you develop your app I think. > I'm looking to sustain about 5,000 msg per second, over 11 hours a day, 7 > days a week. The rate drops substantially outside those hours but is never > zero. > > Mark > > Johnson > Principal Product Architect > > T: +442475269508 > M: 07764305692 > E: mark.john...@flooid.com > Click here to send me something sensitive or securely! > -----Original Message----- > From: Clebert Suconic <clebert.suco...@gmail.com> > Sent: 29 September 2022 01:29 > To: users@activemq.apache.org > Subject: Re: Is Artemis Production Ready? > > *** Warning: This email originated from outside of Flooid’s email system. > DO NOT CLICK LINKS or ATTACHMENTS in this email unless you recognise the > sender and know the content is safe. *** > > It's always nice to hear stories like this... > > valeu e Obrigado! :) (sorry for the PT-br... just saying thanks a lot with > a personal touch) > > On Wed, Sep 28, 2022 at 7:38 PM Edson Richter <edsonrich...@hotmail.com> > wrote: > > > > We actually run Artemis in production with more than 1.000.000 messages > processed per week (evenly distributted during the 24 hours of the day), in > a VM with 64GB RAM, 16 vCore and 256GB SSD NVME in running in raid mode. > > Performance is a quite dificult subject to talk about, because all > depends on where your server run, where your consumers run, what is your > hardware, bare metal or virtual, container or not container, message size, > routing rules, replication rules, protocols, disk controller, disk type and > disk size, etc. In our case, 90% of this 1.000.000 messages have 44 bytes > on length, only 10% are complex json objects. > > > > Few months ago we had few problems (misbehavior), but they got quickly > fixed by downloading source code and fixing it ourselves creating a custom > distro. Few weeks later, Artemis Team fixed the problem and we returned to > the standard distro. > > > > > > Regards, > > > > ER. > > > > > > Enviado do Email<https://go.microsoft.com/fwlink/?LinkId=550986> para > > Windows > > > > De: Clebert Suconic<mailto:clebert.suco...@gmail.com> > > Enviado:quarta-feira, 28 de setembro de 2022 18:50 > > Para: users@activemq.apache.org<mailto:users@activemq.apache.org> > > Assunto: Re: Is Artemis Production Ready? > > > > publishing a benchmark is a game without end. > > > > > > It's always possible to get a particular usecase or tweak things in a > > way that will move the benchmark in any direction you want. > > > > > > the best you can do is to measure the use case you want to achieve > > yourself. I have spent a lot of time with benchmarks before and I > > don't want to get back to that game myself :) > > > > > > Having said that, we have an ./artemis perf client tool as part of the > > artemis cli that you can use for some metrics. > > > > On Wed, Sep 28, 2022 at 4:05 AM Francois Papon > > <francois.pa...@openobject.fr> wrote: > > > > > > Hi, > > > > > > /"the reporting performance of Artemis is significantly higher than > > > Classic"/ > > > > > > I'm very interested about such of reporting performance between > > > Artemis and AMQ. > > > > > > Is it possible to share? > > > > > > Regards, > > > > > > François > > > > > > > > > On 26/09/2022 16:40, Robbie Gemmell wrote: > > > > Couple minor corrections for anyone else reading later.. > > > > > > > > On Mon, 26 Sept 2022 at 14:15, Clebert Suconic > > > > <clebert.suco...@gmail.com> wrote: > > > >> the major bit from the release (2) only tells you about the API. > > > >> Currently version 2 will be version 2 as long as we keep the API > > > >> compatible with previous releases. (When we make it 3.0 it means > > > >> we can remove a few deprecated methods and other stuff) > > > >> > > > >> > > > >> The second bit, 2.26.0 (26), means we had ** Twenty Six ** > > > >> releases fixing bugs and improvements since we released the very > first 2.0 back in 2017: > > > >> https://github.com/apache/activemq-artemis/releases/tag/2.0.0 > > > >> > > > > Due to 2.y.z releases as well, it is actually now 34 releases since > 2.0.0. > > > > > > > >> ActiveMQ Artemis was initially donated from HornetQ, and back > > > >> then we made a roadmap for features we must implement to get the > > > >> same features from ActiveMQ. I believe at this point we are > > > >> already beyond.. and that page needs some updating probably to > reflect the current state. > > > >> > > > >> Also, to talk about production ready quality, the codebase of > > > >> ActiveMQ Artemis was donated to ActiveMQ back in 2017 from > > > >> HornetQ. It is a very stable codebase. I have myself dedicated > > > >> the past 14 years of my profession to this codebase... along > > > >> other developers who I highly consider, and many other open source > contributors... So it is definitely production quality. > > > > It was late 2014 for the donation, 2015 for the Artemis 1.0.0 > > > > release, and then 2017 had the Artemis 2.0.0 release. > > > > > > > >> Talking about that, I'm releasing 2.26.0 today. > > > >> > > > >> On Mon, Sep 26, 2022 at 6:12 AM Mark > > > >> Johnson<mark.john...@flooid.com> > > > >> wrote: > > > >> > > > >>> Although Artemis is at Release 2, I cannot find a direct > > > >>> statement in the online documentation that Artemis is production > > > >>> ready. In contrast, this page suggests that Artemis is *not* > > > >>> production ready > https://activemq.apache.org/activemq-artemis-roadmap. > > > >>> > > > >>> > > > >>> > > > >>> Naturally, I must provide evidence that Artemis is considered > > > >>> production ready by the ActiveMQ team before investing any > > > >>> further effort in deploying and testing Artemis to replace Classic. > > > >>> > > > >>> > > > >>> > > > >>> We are considering Artemis simply because the reporting > > > >>> performance of Artemis is significantly higher than Classic. > > > >>> > > > >>> > > > >>> > > > >>> > > > >>> > > > >>> * Mark* > > > >>> > > > >>> > > > >>> > > > >>> *Johnson* > > > >>> > > > >>> Principal Product Architect > > > >>> > > > >>> > > > >>> > > > >>> Flooid, PCMS House, Torwood Close > > > >>> > > > >>> Westwood Business Park > > > >>> > > > >>> Coventry, CV4 8HX, United Kingdom > > > >>> > > > >>> T: +442475269508 > > > >>> > > > >>> M: 07764305692 > > > >>> > > > >>> E:mark.john...@flooid.com > > > >>> > > > >>> * flooid.com<https://www.flooid.com/>* > > > >>> > > > >>> *Click here to send me something sensitive or securely! > > > >>> <https://sendsafely.pcmsgroup.com/u/Mark.Johnson%40flooid.com>* > > > >>> > > > >>> [image: Download now] > > > >>> <https://www.incisiv.com/playbook-unified-commerce-for-grocery-c > > > >>> onvenience-retail> > > > >>> > > > >>> The information contained in this e-mail is intended only for > > > >>> the person or entity to which it is addressed and may contain > > > >>> confidential and/or privileged material. If you are not the > > > >>> intended recipient of this e-mail, the use of this information > > > >>> or any disclosure, copying or distribution is prohibited and may > > > >>> be unlawful. If you received this in error, please contact the > > > >>> sender and delete the material from any computer. The views > > > >>> expressed in this e-mail may not necessarily be the views of > Flooid Ltd and should not be taken as authority to carry out any > instruction contained. > > > >>> Flooid Ltd reserves the right to monitor and examine the content > > > >>> of all e-mails. Flooid Ltd is a company registered in England > > > >>> and Wales with company number 1459419 whose registered office is > > > >>> at PCMS House, Torwood Close, Westwood Business Park, Coventry > > > >>> CV4 8HX, United Kingdom. VAT No: GB 705338743. > > > >>> > > > >>> > > > >>> > > > >> > > > >> -- > > > >> Clebert Suconic > > > > > > > > -- > > Clebert Suconic > > > > > -- > Clebert Suconic > -- Clebert Suconic
