Conversation

Notices

1. Embed this notice
   GeePawHill (geepawhill@mastodon.social)'s status on Saturday, 15-Jun-2024 02:55:44 JST GeePawHill

   Me talking to people worried about scaling:

   "You're right, we're gonna need a 2nd generation FitBit w/an external drive to handle that kinda load."

   • clacke likes this.
   • Embed this notice
     Chris Siebenmann (cks@mastodon.social)'s status on Saturday, 15-Jun-2024 02:55:42 JST Chris Siebenmann

     in reply to

     • Ben Ford :grinchsmile:

     @binford2k @GeePawHill My techblog uses a CGI that runs a giant Python program (with some efficiency hacks some of the time), on a shared VM that has a total of 4 GB of RAM and a not that modern CPU. It has repeatedly made the front page of that orange site and a few other places, and basically no one has noticed.

     It isn't the machines that are slow, it's the software. (And my other rant is "you don't need static HTML to stand up to loads", although static HTML makes it much, much easier.)

     Haelwenn /элвэн/ :triskell: likes this.
   • Embed this notice
     GeePawHill (geepawhill@mastodon.social)'s status on Saturday, 15-Jun-2024 02:55:43 JST GeePawHill

     in reply to

     The number of teams I've met who are doing AWS and/or k8s for a service mesh to handle loads a single 1985 PC could deliver, smdh.

     25k users, 100 transactions each a day.

     Baby, *any* monolith could deliver that, if your team doesn't start with the assumption that http is free.

     Haelwenn /элвэн/ :triskell: and clacke like this.
   • Embed this notice
     Ben Ford :grinchsmile: (binford2k@hachyderm.io)'s status on Saturday, 15-Jun-2024 02:55:43 JST Ben Ford :grinchsmile:

     in reply to

     @GeePawHill my LUG webserver at university got slashdotted, hit Reddit, and hacker news at the same time in 2008 or so. Oh, and k5 too, RIP. It was a bog standard off the shelf PC tucked away in an office, running drupal on postgresql. We consumed 99.9% of the university’s traffic and got a nice little visit from IT the next day.

     And the machine did indeed slow down a little bit. I mean, if you were looking for it.

     I think about that experience a lot these days when people talk about scale.

     Haelwenn /элвэн/ :triskell: likes this.
   • Embed this notice
     GeePawHill (geepawhill@mastodon.social)'s status on Saturday, 15-Jun-2024 02:55:44 JST GeePawHill

     in reply to

     "5000 users at 5 transactions a day!?!? Do you have a 'smart' refrigerator?"

   • Embed this notice
     GeePawHill (geepawhill@mastodon.social)'s status on Saturday, 15-Jun-2024 02:55:44 JST GeePawHill

     in reply to

     "Well, to be honest, the single 8gb VPS w/the 2tb disk *is* gonna run you. You're looking at maybe $50/month."

   • Embed this notice
     GeePawHill (geepawhill@mastodon.social)'s status on Saturday, 15-Jun-2024 02:55:44 JST GeePawHill

     in reply to

     I know it's a very unpopular old-school stance -- I'm 64yo, and I'm aging out -- but I think you maybe should make some money off your first 25k users hitting a monolith on a VPS before you start paying AWS bills.

   • Embed this notice
     GeePawHill (geepawhill@mastodon.social)'s status on Saturday, 15-Jun-2024 02:57:19 JST GeePawHill

     in reply to

     The powerless I speak this truth to, almost uniformly, give me the old-man-yells-at-cloud look.

     Haelwenn /элвэн/ :triskell: likes this.
   • Embed this notice
     clacke (clacke@libranet.de)'s status on Saturday, 15-Jun-2024 23:44:04 JST clacke

     in reply to

     • Andreas Sikkema
     @bix @GeePawHill My historical perspective is so bad, things have shifted so fast ... 256 MiB was really server-sized RAM 20 years ago?
   • Embed this notice
     Andreas Sikkema (bix@chaos.social)'s status on Saturday, 15-Jun-2024 23:44:06 JST Andreas Sikkema

     in reply to

     @GeePawHill in 2001 or 2002 a colleague of mine showed our proprietary VoIP PBX software running on a Toshiba Libretto (486 iirc running Windows 2000) and the rest of the company never understood the power of that, because servers are 19” pizza boxes with quad Itanium processors and 256 MByte RAM…

   • Embed this notice
     nick (nickzoic@aus.social)'s status on Saturday, 15-Jun-2024 23:45:08 JST nick

     in reply to

     @GeePawHill yeah, a conversation I have quite frequently

     "okay roughly how many transactions per second are we talking?"

     "we don't know."

     "okay let's work that out so we can plan ..."

     "we don't need to because we will implement on k8s using cutting edge caching technologies and service network architectures and therefore everything will be web scale"

     "... okay I come up with 5 tps so allowing for ... wait, what?"

     clacke likes this.
   • Embed this notice
     Joshua M. Clulow (jmc@m.unix.house)'s status on Sunday, 16-Jun-2024 14:49:45 JST Joshua M. Clulow

     in reply to

     @GeePawHill Definitely a big fan of small monoliths and SQLite and single VMs. Admittedly in AWS because it's been easy, but I am using almost no features of EC2 other than "a small VM please".

     clacke likes this.
   • Embed this notice
     jzakotnik (jzakotnik@mastodon.social)'s status on Sunday, 16-Jun-2024 14:49:51 JST jzakotnik

     in reply to

     @GeePawHill is performance/load the main argument why they use k8s? My impression is that the value comes with more flexibility in the development, eg I can easily automate pipelines for any number of feature branches etc. That used to be tricky without a good container platform/cloud.

     clacke likes this.
   • Embed this notice
     clacke (clacke@libranet.de)'s status on Monday, 17-Jun-2024 00:56:06 JST clacke

     in reply to

     @GeePawHill Is performance why they're doing it though? I'd do it because a mostly idle pod is cheaper than a VM is cheaper than any kind of physical space, plus how easy it is to have a proper deployment pipeline and zero downtime on redeployment.

     Not simpler, far more complex and lots of moving parts. But easier.