Search results

No one actually cares about the laser itself damaging anything. Every non-esoteric optical module is a class 1 laser under the hood, and they need no special protection. Try not to start into them all day; the boredom isn't good for you. The problem is that the optics can fail (or be damaged)...

I'm kind of expecting combo PoE + SFP to be increasingly common as WiFi speeds keep climbing. There's no way that anything faster than 10G over copper is going to happen any time soon. And even if someone *does* dust off the never-commerically-implemented 40Gbase-T spec, then it'd still need...

UniFi takes heat from "real" network people, but it's probably pretty reasonable for this case. It's simple, it's fairly cheap, and it works reasonably well. It's not fancy, it's not all that flexible, and it'll probably fail earlier than better-built enterprise hardware, but I doubt any of...

One generic recommendation: whenever possible, prefer singlemode fiber over copper. It's not always possible (you can't run PoE over fiber, for example), but it's a much better choice for running between switches or anywhere else that doesn't *have* to be copper, for a few reasons: It's...

You can also ask SNMP about how much unicast/multicast/broadcast traffic is going out each of the switch ports. If you start ~1 Gbps of traffic between your PC and NAS and you see ~1 Gbps of traffic *also* hit your WiFi uplink ports, then that's a pretty good indication of where your problem...

It's unlikely, but I had something similar once that turned out to be an Ethernet driver bug on my file server; all of the outbound Ethernet frames were being sent to 00:00:00:00:00:00, the broadcast MAC address. So all of the NFS, etc traffic was being spammed to every host over wired or WiFi...

Nice to see that it works. I'm kind of surprised that it was only a ~6% win on memory latency given the ~15% change in DIMM timings, but I've never really paid a ton of attention to benchmarks and RAM performance. Presumably that's normal?

I keep hoping that 1x100G is the next 10G/25G. It's roughly similar in performance to a x4 PCIe 4 NVMe, which should mean that it's fast enough for nearly all small-scale storage needs. Mind you, even 50G lanes are pretty rare right now, and I'm not convinced that there's a lot of...

FWIW, the 7050QX-32S can have its fan speed dropped pretty substantially via a config setting. At that point it's merely annoyingly loud. I have 2 in a closet close to my home office, and it's just fine, now that I've turned down the fans, installed sound baffling tiles on the inside and...

Doing 10Gbase-T requires a lot of processing with a hefty DSP, and this historically used 2-4W of power. The newest modules are a bit better, but there's still a lot of work involved, and the newest/lowest power/coolest ones are usually the most expensive. I'm using a MS-01, but I'm using a...

ConnectX cards probably won't do 2.5G*. They *really* want SFPs, and they usually need active cooling or they'll heat up to over 120C and shut down within a few minutes. Nice cards, but overkill for 2.5G. * - You could probably make it work with one of the few 10Gbase-T SFP+s that can talk...

One minor thing to keep in mind is that LGA 2011 can be either PCIe 2 or PCIe 3, depending on the CPU generation. So, sticking a new PCIe 4.0 x1 NIC into a slot would cap at at around 2 Gbps with an E5 v1 (Sandy Bridge) or 4 Gbps with an E5 v2, v3, or v4 CPU. Intel i226 NICs would be best; I...

*At best*, modern HDDs will get around 250 MB/sec, but you're unlikely to see that under real world use, and a lot of access patterns can drop that down under 1 MB/sec. It all depends on your files and what you're doing with them. Remember that HDDs can really only do ~100 seeks per second...

Since it sounds like you're kind of new to networking, I figure it's not a bad idea to throw in a few guidelines. In general, there's no real downside to throwing a small unmanaged switch or two in the middle of a home network, as long as you remember a few simple things: Make sure that you...

I mean, it seems like 1/3rd of last year's STH articles were reviews of small 2.5G switches. They've shifted a bit more towards 5G/10G lately, but there are still quite a few of them. Here is the Amazon link from the most recent one...

That plus a tiny udev hack is enough to have 65 usable interfaces (64x QSFP28 plus 1 management port) working with VyOS. Now I can move this beast away from my desk and start using it for VyOS development in earnest.

The big problem is that `mstfwmanager` isn't installed as part of VyOS, and it's not compiled into the Debian bookworm `mstflint` .deb. I don't have a bookworm build system handy, either. In the end, I just dropped a bookworm APT config onto the switch and downloaded GCC and friends and built...

Hey, I have a SN3800 that I'd like to use with a stock Linux distribution (using it for VyOS dev), but it's currently running too old of an OS (3.8.2204) for the Linux `mlxsw` driver to be able to upgrade it: [ 30.449077] mlxsw_spectrum2 0000:06:00.0: The firmware version 29.2000.2698 is...

The 802.3an standard for 10Gbase-T is from 2006, vs XP from 2001-2007. It doesn't look like hardware shipped before the standard was ratified (unlike a lot of WiFi gear). Solarflare drove a lot of the early hardware, so their NICs are probably your best bet.

This probably isn't *quite* right (it's 8-pin CPU instead of 8-pin PCIe), but they shouldn't be too hard to find: 8 Pin CPU Female to 2 x MicroHi 4 Pin for Supermicro Backplane 50cm It's literally just 2 12V wires and 2 0V wires into a common connector.