Notices by thurloat 📌 (thurloat@mastodon.club), page 48

i t i s w e d n e s d a y m y d u d e s

@lain is it behind a flag, or is no one using it? i'm on 65, and still see my beautiful scrollbars everywhere.

@catalina uh oh, whats happenin

@manu heh, just my 2 year old :)

thank you for your time.

love Ezra.

, llollllllllllllllllllllllllll.lllllllllllllllllllkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkk kkkkkkkkkkkkkkkkkkkkkkm kk kk llllllllllllllllk

bv i888

t g g v n mumuk8k8k

bny df b vf nhj b

m h gnthn j mimi}
.;kmk nm,k.l i0pklo0;k/kl';[pnmhju, l,,lkjuuty5t666yt

@mpjgregoire yea the procurement system is set up pretty strangely.

my sources close to the mil procurement process say it's equally as frustrating and drawn out as the public sees it, they wish they could pull the trigger on stuff without all the bureaucracy.

@yukiame of course! not arguing virtualization is bad.

The difference between 100 customers experiencing degraded performance, and 20 is important. When you extrapolate it, say 5% of your 10,000 customers run intensive VMs. If the design is such that your 5% of customers negatively impact 10% an additional of customers, vs. 2-5% -- that is substantial.

And can't huddle all the "bad" VMs on a few nodes, because they're likely high paying and you'd have an exodus.

@selea yea I know some folks that run Nutanix stuff. They've generally had pretty good experiences.

Mostly because of the way we run our business, hyperconverged has never really been an option once we grew up.

In 2013ish we actually had a HCI-ish design for compute and storage, we aptly named that time period in the company a "GlusterFuck".

@yukiame intelligent hypervisors aren't going to make business decisions for you.

designing compute infrastructure to have a 100VM blast radius for a single node is going to force you into facing the trolley problem at a larger scale. Minimizing the blast radius of any failure is critical, and smashing performance / capacity limits are a design failure.

@selea hopefully at that point they have some SAN cluster or NVMe fabric storage.

@feld they've got thermal density to take care of with increased power limit as well, if you're getting more then they're borrowing from someone else if its not already packed like sardine tins in there.

obviously depends how seriously the facility takes the limits of their design.

@scarlett anti-socks software users

@scarlett is this a thing?

or did i get memed?

people who put 3TB of ram and dual 32 core procs in a 1U virtualization server expecting to host 100 VMs, but don't realize there is such a thing as memory bandwidth.

one or two intensive VMs, and everyone is having a bad time.

then your 40U colo rack in the DC can only have 10U populated because you've hit your power limit.