The OFFICIAL tech stuff thread

Kilemall

@madrebel said in The OFFICIAL tech stuff thread:

@oyaji said in The OFFICIAL tech stuff thread:

or you could just pay… and pay… and pay…

Let me Steel Man the argument from the car OEM side.

“We’re already wiring up all the features whether you use them or not, our plan then is to deliver one base model that you can ‘enable’ individual features if you like”

That is my assumption from their part. It is easy to be cynical and say this is a money grab, however, you have to consider the supply chain optimization having a single model will deliver for them.

Now between you and I, so not a fan of these models. But I’m not blind to the efficiency this could enable and the fact that this ‘could’ (likely wouldn’t) deliver lower costs for me the end consumer.

Also keep in mind all the car OEMs have very popular leasing models. “Leasing features” is a logical progression IMO.

IBM mainframes literally have had this model for decades.

When you order one, you get one of a few models, all decked out with full capacity cause it’s cheaper to make them all the same.

You then pay by how many MIPS you want (old school term, the current usage is MSUs). IBM engineers then use the ‘golden screwdriver’ to allow your machine to make just the paid for resources available. You can then play around with the settings, including how many CPUs you use, and pay for more I/O channels.

Then the software people get into the act with pricing models based on the MIPS/MSUs. Last time I was involved, the IBM hardware and OS pricing was comparable to their UNIX machines but the software apps people were exorbitant.

Gators1

NGINX Explained - What is Nginx

I’ve been doing this shit professionally for 3 decades and I shake my head at how we can almost max out RAM on a system with FIVE TERABYTES of the stuff because some dude wanted to see what an option in a report did.

Gators1

I think it’s .5 TB unless metric TBs are smaller or something?

Stuff like that is one of the banes of my existence. We do a lot of work to avoid data moving to a user system unless they really need it. Everything else renders the results on the web based on summarized results in the back end or whatever.

Whoofe

@Hog said in The OFFICIAL tech stuff thread:

I’ve been doing this shit professionally for 3 decades and I shake my head at how we can almost max out RAM on a system with FIVE TERABYTES of the stuff because some dude wanted to see what an option in a report did.

but did they use a cover sheet for their TPS report?

@Gators1 said in The OFFICIAL tech stuff thread:

I think it’s .5 TB unless metric TBs are smaller or something?

Machine has 5 terabytes of RAM which the dude almost maxed out by consuming 10% of it (or nearly all of the free memory) for his report.

Gustaf

I saw your poor math prior to your edit!

Gators1

@Hog said in The OFFICIAL tech stuff thread:

@Gators1 said in The OFFICIAL tech stuff thread:

I think it’s .5 TB unless metric TBs are smaller or something?

Machine has 5 terabytes of RAM which the dude almost maxed out by consuming 10% of it (or nearly all of the free memory) for his report.

Sounds like a poorly coded application on the machine if it uses 4.5TB of memory just to run it and leaves the poor user only 500GB for doing actual work with the tool. I would get rid of my offshore contracted programmers if I were them.

oyaji

i remember being told that “no matter how much RAM you have installed, windows will use half or more of it”.

@Gators1 said in The OFFICIAL tech stuff thread:

@Hog said in The OFFICIAL tech stuff thread:

@Gators1 said in The OFFICIAL tech stuff thread:

I think it’s .5 TB unless metric TBs are smaller or something?

Machine has 5 terabytes of RAM which the dude almost maxed out by consuming 10% of it (or nearly all of the free memory) for his report.

Sounds like a poorly coded application on the machine if it uses 4.5TB of memory just to run it and leaves the poor user only 500GB for doing actual work with the tool. I would get rid of my offshore contracted programmers if I were them.

The machine is used by 10K people for hundreds of different processes but, yeah, something’s fucked if one guy came in over the top and consumed 500gb just by running a single report.

I don’t know what the issue is but I’ve been forming the opinion that them running the BI tool on the same hardware as the ERP is a mistake. But I don’t know; it’s all above my paygrade and I don’t have enough info to even make informed speculation. I only even hear about it when it’s close to hitting the fan like today.

Gustaf

Maybe all that RAM is being used by your anti-Cambodian spy services! Sounds like another meeting from the authorities is in order!

madrebel

@Hog said in The OFFICIAL tech stuff thread:

that them running the BI tool on the same hardware as the ERP is a mistake.

you think? two extremely memory hungry applications that also pound the same disk subsytem with random IO on the physical box is an extremely bad idea.

Gators1

@Hog said in The OFFICIAL tech stuff thread:

@Gators1 said in The OFFICIAL tech stuff thread:

@Hog said in The OFFICIAL tech stuff thread:

@Gators1 said in The OFFICIAL tech stuff thread:

I think it’s .5 TB unless metric TBs are smaller or something?

Machine has 5 terabytes of RAM which the dude almost maxed out by consuming 10% of it (or nearly all of the free memory) for his report.

Sounds like a poorly coded application on the machine if it uses 4.5TB of memory just to run it and leaves the poor user only 500GB for doing actual work with the tool. I would get rid of my offshore contracted programmers if I were them.

The machine is used by 10K people for hundreds of different processes but, yeah, something’s fucked if one guy came in over the top and consumed 500gb just by running a single report.

I don’t know what the issue is but I’ve been forming the opinion that them running the BI tool on the same hardware as the ERP is a mistake. But I don’t know; it’s all above my paygrade and I don’t have enough info to even make informed speculation. I only even hear about it when it’s close to hitting the fan like today.

Maybe a cache option is on and shouldn’t be? Typically the db should be streaming the answer, nor storing it… Installing BI on the ERP server is even beyond my company stupid.

Kilemall

@Hog said in The OFFICIAL tech stuff thread:

I’ve been doing this shit professionally for 3 decades and I shake my head at how we can almost max out RAM on a system with FIVE TERABYTES of the stuff because some dude wanted to see what an option in a report did.

Hardware explosion makes for sloppy children.

Kilemall

@madrebel said in The OFFICIAL tech stuff thread:

@Hog said in The OFFICIAL tech stuff thread:

that them running the BI tool on the same hardware as the ERP is a mistake.

you think? two extremely memory hungry applications that also pound the same disk subsytem with random IO on the physical box is an extremely bad idea.

Adding on network bottleneck and overhead is better?

For our bigass apps, we’ll often have the main app where the updates occur and then a shadow reporting database mirrored to the main one just for splitting resources. And then in the most crucial case a nightly download to a third reporting database…

Gators1

@Kilemall said in The OFFICIAL tech stuff thread:

@Hog said in The OFFICIAL tech stuff thread:

I’ve been doing this shit professionally for 3 decades and I shake my head at how we can almost max out RAM on a system with FIVE TERABYTES of the stuff because some dude wanted to see what an option in a report did.

Hardware explosion makes for sloppy children.

Exploding hardware does look cool though.

Gators1

@Kilemall said in The OFFICIAL tech stuff thread:

@madrebel said in The OFFICIAL tech stuff thread:

@Hog said in The OFFICIAL tech stuff thread:

that them running the BI tool on the same hardware as the ERP is a mistake.

you think? two extremely memory hungry applications that also pound the same disk subsytem with random IO on the physical box is an extremely bad idea.

Adding on network bottleneck and overhead is better?

For our bigass apps, we’ll often have the main app where the updates occur and then a shadow reporting database mirrored to the main one just for splitting resources. And then in the most crucial case a nightly download to a third reporting database…

Same, we have a replication layer for reporting/extraction and then extract data from that into a lake for BI.

Kilemall

@Gators1 said in The OFFICIAL tech stuff thread:

@Kilemall said in The OFFICIAL tech stuff thread:

@madrebel said in The OFFICIAL tech stuff thread:

@Hog said in The OFFICIAL tech stuff thread:

that them running the BI tool on the same hardware as the ERP is a mistake.

you think? two extremely memory hungry applications that also pound the same disk subsytem with random IO on the physical box is an extremely bad idea.

Adding on network bottleneck and overhead is better?

For our bigass apps, we’ll often have the main app where the updates occur and then a shadow reporting database mirrored to the main one just for splitting resources. And then in the most crucial case a nightly download to a third reporting database…

Same, we have a replication layer for reporting/extraction and then extract data from that into a lake for BI.

BI, the most common data kingdom out there.

madrebel

@Kilemall said in The OFFICIAL tech stuff thread:

Adding on network bottleneck and overhead is better?

network bottleneck? oh you must be poor and can’t afford 100GbE.

Kilemall

@madrebel said in The OFFICIAL tech stuff thread:

@Kilemall said in The OFFICIAL tech stuff thread:

Adding on network bottleneck and overhead is better?

network bottleneck? oh you must be poor and can’t afford 100GbE.

Overhead on feeding it in and handling while not particularly reducing memory/IO, just keeping the query hog itself off.

And I already told you the solution so piss off with your alpha shit.