Recently sat through a TCO by Broadcom showing us how much they're gonna save us by going full VCF across the environment. Apparently my 2 dedicated storage guys make $3.2mil/year. What an incredible savings! 
what did you learn today? (part 2)
- Thread starter continuum
- Start date
I can't even comprehend what kind of storage $3.2m/year gets you.Recently sat through a TCO by Broadcom showing us how much they're gonna save us by going full VCF across the environment. Apparently my 2 dedicated storage guys make $3.2mil/year. What an incredible savings!
I'm feeling kind of underpaid right now. Imagine getting paid $1.6m/year to babysit some storage.
I already asked my VP if I could step down from architect into just storage pleb. He said no 
Thicker copper should be more efficient, not less, because the resistance is lower.
Sure, but it costs at least proportionally more to procure, and is maybe more labor to install.
Pricing is bad enough that we are quickly moving away from our remaining in-office VMware ESX hosts. We renewed last year for 1 year because we had planned to reduce our compute costs in offices due to business needs. (centralization, SaaS, etc.). They then told us we couldn't reduce our core count, and that they weren't going to help us move away from VMware. We weren't looking to move away, we just had reduced based on our company requirements these days. They wanted us to buy 6x more licensing than we needed.
The guy was being a real asshole, so we will now be fully off VMWare in our offices in the next 2 months. They threatened us saying that we couldn't run unlicensed, we would be out of compliance, etc. No worries - we will be fully off. BYEEEEEEEE!
Last edited:
If you didn't convert your licenses to subscription, you won't be out of compliance by not paying them any more. You won't get patches/etc, but you won't be out of compliance. If you've converted to subscription, you can no longer use that software at all after your subscription runs out. Just for anyone that gets the same BS from their broadcom schmucks.
It's driven at least one of our clients who had a dedicated environment onto our shared environment.
We're .. coping, I think is the term I want to use. It's a dead end for sure in the next three to five years, even more than it was pre-Broadcom acquisition.
I am not really surprised. We are accelerating the move of one datacentre to Azure. The ones on the ships will probably get something not vmware next replacement cycle.
I've played with xcp-ng and proxmox at home and man I would never want to use any of that at work.
I've got a USB hard drive passed through to a VM (plex) and randomly on a reboot it decides that device shouldn't be reattached. So that's fun.
I've got a USB hard drive passed through to a VM (plex) and randomly on a reboot it decides that device shouldn't be reattached. So that's fun.
If you pass it through as /dev/disk/by-id, that shouldn't normally happen. I'm doing passthrough here on a proxmox box and it's never failed to come up correctly.
We’re running mostly xcp-ng with Orchestrator. It’s not vSphere, but it’s good enough for us for now. And a damn site cheaper than paying those fucks at Broadcom.
Its passed through via the web admin thing (XOA).
The drive doesn't go away, xcp decides it should no longer be perpetually mapped. Other people have reported it as an issue.
When I did a drive passthrough in Proxmox, it took a CLI command to do it properly. The web interface didn't work right, for some reason. I don't think you can pass by disk-id with the GUI. After doing it with the CLI, it's been flawless.
I don't remember the syntax, but it was pretty easy. The hardest part was finding the disk id. I can figure it out again if you're stuck.
Sure, but when you reduce voltage and increase amperage, resistive losses go up. So you spend more money on thicker copper just to break even on efficiency.
But you're not reducing voltage. You're reducing resistance by increasing copper thickness. Which means, in other words, you're fractionally increasing voltage, by reducing cable loss.
By your argument, the thicker the cable, the worse the problem would be. At some eventual thickness, it would be glowing red hot.
Yeah, I don’t think I said that very correctly. If you’re going from 208V to 48V, you’ve got to carry more current, and thus need more copper to do so. Power loss = Current Squared * Conductor resistance. So if current goes up by a factor of 4, you need 16 times as much copper to maintain the level of efficiency. But I don’t know anything about the reality of distributing 48V DC power, so there’s probably something I’m missing.
Also, aren't you crossing 2 completely different things here? 208V AC to 48V DC? Doesn't that also factor in with your current loads and resistance factors when comparing wiring- let alone the Ampreage draw? It's almost like comparing a diesel hauler truck to a gasoline pickup. Not exactly an apples to apples comparison.
48V architecture is replacing multiple phases and inverter stacks by an in rack version. This is already a thing at hyperscale DCs:
For medium density systems, 36 kW is 240V 150A, still lower than those silly Nvidia DGX racks that hit triple digit kW ratings.
For medium density systems, 36 kW is 240V 150A, still lower than those silly Nvidia DGX racks that hit triple digit kW ratings.
T Mobile's network seems to be messing with VPN traffic when you tether a laptop to your phone.
As early as the first Android (G1) I can remember people working on hacks because TMobile has always treated tethered data different than phone data. Not shocking they're intentionally or unintentionally breaking something.
Going through the CIS 2019 Baseline PDF I feel lke I am reading one originally meant for 2008 by the numerous references to things breaking on NT4 systems or Win9x systems
It seems to come and go, which is the annoying part.
The real fix would be cradlepoints in their service vehicles and a private APN but that conversation doesn't seem to get any traction.
More than normal?
I know a lot of mobile networks seem to have a practical MTU in the 1200s
Yeah, we weren’t even seeing the packets hit the firewall.
IPSec or TLS?
You'll probbly get better luck over V6, it is Tmobile after all.
You'll probbly get better luck over V6, it is Tmobile after all.
IPSec. I've got an order in to get them an IPv6 block provisioned to see if it helps.
I do so very much love having equipment in cheap colo spaces and getting bit on the ass (again) by 120v power.
I so very much love having one single device in a server room that is incapable of taking 208/240v power like literally everything else.
I'm looking at you Adtran. Their SBCs (at least the 908e) with PSUs integrated to the system PCB that are strictly rated for 120v.
The magic smoke has been let out of at least two because of that nonsense.
This means both a step-down transformer off the room UPS, and a separate switched PDU just for that one device. Oi
I'm looking at you Adtran. Their SBCs (at least the 908e) with PSUs integrated to the system PCB that are strictly rated for 120v.
The magic smoke has been let out of at least two because of that nonsense.
This means both a step-down transformer off the room UPS, and a separate switched PDU just for that one device. Oi
Last edited:
I'm relatively new to this whole server world, but I learned from my school's IT manager that IaC is an incredibly useful concept. Still a little worried about using Terraform for my personal project, but I'm up for the challenge.
We've used APC AP9627's for this, not a cheap date.
I'm a little unsure of this - the issue is that the equipment can't handle inconsistent 120v where 208v(?) would have more redundancy?
Nah, it's more that cheap colos (including the one we're using) only have 120v power to the racks. Which is a problem when you're looking to upgrade the storage running the apps hosted out of there and Dell is telling you you'll need 208 or 240v power to run the storage they're quoting you.
