FS.COM is pretty good. I used to work at a place that used 10baseFL. It was a big waste of money. Every computer that came in had an Intel 100mbps nic, it got yanked and replaced with a $400 10 megabit fiber card. But when lightning struck near one of the buildings, it blew out like $100K worth of Lucent Definity phone boards and every HP cat5 jetdirect board. Now at the $day_job at my best point it's a quad LAG of 100-gigabit, for 400 gigabit total. How far things have come. At last job I saw a rack half-full roughly sustaining a terabit per second to the internet. 12 x 1u servers maybe. I still have a Sun and Indigio2 FDDI card or two somewhere, and one for Indy. At one point in time I had like 5 Indy/Challenge S all on FDDI. Will probably bring some of it to the next swap event. - Ethan On Mon, 15 Feb 2021, Jameel Akari via vcf-midatlantic wrote:

Before somebody says "it's not vintage!" consider that 1Gb Ethernet on fiber was spec'd for FDDI cables, and there were optical transceivers for even 10Mb Ethernet - I brought a set of AUI to 2xST for the Unix Town networking, just because I had them. Also, early 1Gb Ethernet was all 1000base-SX with SC connectors, at least that I saw. It took a while for copper UTP to be supported. I have a few Digital and Sun 1Gb NICs that are optical, and so I run them that way.

Now, at 25m, you /really/ don't need fiber for 1Gb Ethernet. Properly installed UTP copper should work reliably for years in normal environments. It's surprising and unusual you had some fail. Exception: crimped RJ45 ends on solid conductor UTP cable is likely to fail over time if there's any movement or vibration. Punch on some jacks and use premade jumpers.

You can run 10Gb over Cat6A UTP at this distance as well. This is harder to terminate by hand, but you could just get premade patch cables (they will be 26-28 AWG stranded, most likely)

But since you've already got optics and fiber is a lot cheaper than it used to be... why not?

- You're better off not trying to use the same cabling for 1Gb and 10Gb, but: - You *can* do short-range 1000base-SX with OM3, though it's not spec depending on where you look. It will work. - By the book, use OM2 for 1Gb and OM3 for 10Gb, assuming short-range optics like your SX SFPs. (SFP, not GBIC, if you're talking LC connectors) - OM2 is usually orange and OM3 is usually aqua-blue.

- As far as I can tell, you don't need mode conditioning for SX (short-range) optics. If you needed to use LR 1Gb optics, or use single-mode optics into MMF cabling plant, then you'd use a mode-conditioning cable to clean up the noise.

- Bend radius is typically 10x the outside diameter of the cable, unless otherwise specified. Here, it's a single strand of a duplex cable. So if it's about 3mm across, bend radius is 30mm. (but see below on bend-insensitive cable). When I dress cables I usually eyeball 3-4" diameter loops. Any smaller just doesn't look or feel right anyway.

I'm assuming that at only 25m you're buying pre-terminated cables with LC plugs both ends. This means you need large enough holes in things for those plugs to fit through. Otherwise it's definitely not worth splicing your own at this scale.

https://www.fs.com/products/74386.html?attribute=226&id=99219 - for $16.72 each at 25m, you get new "bend insensitive" cable (7.5mm radius!)

I love FS.com for all manner of fiber and related cabling: https://community.fs.com/blog/advantages-and-disadvantages-of-multimode-fibe... https://community.fs.com/blog/single-mode-cabling-cost-vs-multimode-cabling-... These have some rule-of-thumb charts.

I think that covers it?

-- Jameel Akari

On 2021-02-15 12:25, Neil Cherry via vcf-midatlantic wrote:

...

I've posted this in the CDL list also. I don't know of a polite way to post once to both and I'm not sure it's a good idea to try. :-)

It has been a while since I played with fiber (2010) and I'm trying to figure out if I should go fiber or copper in my home. Here's what I know from my current switch:

- SX GBIC - LC - duplex - MMF

I want to buy about 25M of fiber. I think the price is competitive with copper. I'm also hoping that the fiber will age better than the copper. I'm seeing a few old copper drops fail to lower settings. One went to 10/half (wow, that's bad).

The GBIC & switch I have handle 1G. So I can easily pop in a few 1000baseT, put them into a LAG and get 4G (4 ports). So I'm covered for a while.

- Can the same piece of fiber handle 1G or 10G? (in case I upgrade in the future). - Does anyone know the bend radius of this fiber? - What color should I expect for SX fiber. ? - What's a "mode-conditioning patch cord" ?

I'm seeing a $35 price for 25 Meter OM3 LC LC Fiber Patch Cable (Corning 50/125um core/cladding) and it says 1G but with a "mode-conditioning patch cord" (an attenuator?).

Any pointers?

On 2/15/21 3:06 PM, Jameel Akari via vcf-midatlantic wrote: Thanks Jameel, Glad you didn't quash my dreams. :-) I think I'm going to go UTP for now. I can toss 4 in a LAG and get plenty of bandwidth for now. I will try to remember fs.com. -- Linux Home Automation Neil Cherry ncherry@linuxha.com http://www.linuxha.com/ Main site http://linuxha.blogspot.com/ My HA Blog Author of: Linux Smart Homes For Dummies

If I had a house I would run OM3, I like it better than copper and I have no problem spending the extra money on the SFPs. We have some rather long OM3 runs lit with 1MbpsSX and 10Gig SR SFPs. We get our optics at OSI and I run Juniper switches.

If I had one I would probably run both, cat5 has the advantage for PoE devices. SFPs should be dirt cheap for anything 10gig and below. The problem on the 40 Gigabit and 100 Gigabit stuff (I think Alex is runing 40 Gig at home?) is the normal stuff is all actually 8 fibers in the cable. It really 4 x 10gig or 4 x 25gig in one module. I assume there are some optics with prisms built in that DWDM but probably not $8 on feeBay. I think my 10gig SFPs were $9 each for Intel or fake Intel. My switches usually come from trash piles so the 10 gig ports were SC, but finally moving up to LC.

We ditched 10G in our top of rack because even at 3m we were having issues. We replaced the top of rack from copper to SFP and added intel 10G SFP cards on the servers. We use 10Gbps DAC cables for 20G per server. All of this is of course personal preference.

Interesting. I know some of the switches don't like copper SFPs that do 10 gig, apparently they chew a lot of power. Oddly most of my stuff is short, but all the cables are Direct attach fiber at 25gig and 100gig. The vendor lock in on the SFPs is totally annoying though. - Ethan

.

Oh forgot, Chris my condolences on Juniper ;-) I hated working with their IPv6 configs, trashed a lot of router/switch configs. Still favor Cisco though I'm not sure what a SDN configured device looks like anymore. Labs hit us (mS testing) with PNFs with 4K interface objects. Hmm, SDN is getting hard to describe in anything that's easy to make sense of.

I'll take a Juniper over a Cisco any day, again personal preference. I really like the Juniper CLI and I am very comfortable with their platform. We looked at "SDNs", we run NSX and we just use the firewall. Problem is with complex solutions you need a team that can troubleshoot them, most people can't troubleshoot simple ethernet. So we keep things simple and old school. For example, we don't run QoS, we just invest in bigger pipes.

My current switch can handle 1G (x4, SFP). I don't I'll need more than 4G of traffic between the two switches. Most of it is small, short messaging or streaming (only 2 TVs). Later there will be cameras (POE). But that design work for later.

I don't think you will need more than 4G either. I have an office with 2000+ people and 3x10G internet access. Our users don't peak past 6G. When you finish your setup let me know. I like to see other designs and I like opinions about other network devices.

On 2/16/21 1:34 PM, Ethan O'Toole via vcf-midatlantic wrote:

Pretty crazy thinking how much data they can theoretically move.

I've seen references to 100Gi and 400Gi but I only see topology and it's hard to tell what's real as we can mock up data. A LAG built with 100Gi is pretty wild, 400Gi LAGs seems really farfarfarout. ;-) As far as complexity, how about 'white boxes' where you load on Vendor X's OS on a generic network device and suddenly the network box looks like vendor X despite not even having that hardware in the box. Add SDN and IP over IP and vlans (and vlans in vlans) and I have some really interesting discussions with the SEs. From morse code speed (28,8 bps perhaps) to 400Gi. in a span of ~200 years. -- Linux Home Automation Neil Cherry ncherry@linuxha.com http://www.linuxha.com/ Main site http://linuxha.blogspot.com/ My HA Blog Author of: Linux Smart Homes For Dummies

On 2/17/21 11:21 AM, Alexander Jacocks wrote:

I’m running primarily 10 Gbps infrastructure, using a 24 port Arista fiber switch as my core. I also do have 2 8-port 40 Gbps switches, one ethernet and one Infiniband. Hamfests have been good to me!

Uhm, is this home? I'm not trying to be the one with the greatest toys and if I was I was failing pretty badly. ;-)

I will need to add another 10 Gbps switch, so that I can manage all the new Cat8 runs, but for now, those will likely terminate in a 1 Gbps switch, until I find another good deal.

Cat 8? Good grief. ;-) https://www.cablesandkits.com/learning-center/what-are-cat8-ethernet-cables -- Linux Home Automation Neil Cherry ncherry@linuxha.com http://www.linuxha.com/ Main site http://linuxha.blogspot.com/ My HA Blog Author of: Linux Smart Homes For Dummies

On Mon, 15 Feb 2021, Dave McGuire via vcf-midatlantic wrote:

On 2/15/21 2:38 PM, Ethan O'Toole via vcf-midatlantic wrote:

...

I used to work at a place that used 10baseFL. It was a big waste of money. Every computer that came in had an Intel 100mbps nic, it got yanked and replaced with a $400 10 megabit fiber card. But when lightning struck near one of the buildings, it blew out like $100K worth of Lucent Definity phone boards and every HP cat5 jetdirect board.

I think the "FL" in 10baseFL stands for "FLorida". Lightning capital of the world and all. I lost about half a dozen copper ports down there before I developed a rule for my network: If it leaves the room, it's on fiber.

When I was working in MD, we had a customer with a long RS-232 run between buildings. I was there several times to replace the 1488 and 1489 chips in the remote printer. They never did do anything about it. Mike Loewen mloewen@cpumagic.scol.pa.us Old Technology http://q7.neurotica.com/Oldtech/

On 2/15/21 5:37 PM, Mike Loewen via vcf-midatlantic wrote:

On Mon, 15 Feb 2021, Dave McGuire via vcf-midatlantic wrote:

...

On 2/15/21 2:38 PM, Ethan O'Toole via vcf-midatlantic wrote:

...

 I used to work at a place that used 10baseFL. It was a big waste of money.  Every computer that came in had an Intel 100mbps nic, it got yanked and  replaced with a $400 10 megabit fiber card. But when lightning struck near  one of the buildings, it blew out like $100K worth of Lucent Definity  phone boards and every HP cat5 jetdirect board.

 I think the "FL" in 10baseFL stands for "FLorida".  Lightning capital of the world and all.  I lost about half a dozen copper ports down there before I developed a rule for my network: If it leaves the room, it's on fiber.

   When I was working in MD, we had a customer with a long RS-232 run between buildings. I was there several times to replace the 1488 and 1489 chips in the remote printer. They never did do anything about it.

I worked on boards added to the AP news wire printers (Okidata 80's a later). If someone mentioned lightening I didn't even diagnose the problem. I just popped the two out and tested. They were good chips and sacrificed themselves for the rest of the printer. :-) -- Linux Home Automation Neil Cherry ncherry@linuxha.com http://www.linuxha.com/ Main site http://linuxha.blogspot.com/ My HA Blog Author of: Linux Smart Homes For Dummies

On 2/16/21 10:38 AM, Ethan O'Toole via vcf-midatlantic wrote:

...

  When I was working in MD, we had a customer with a long RS-232 run between buildings. I was there several times to replace the 1488 and 1489 chips in the remote printer. They never did do anything about it.

As long as the ICs are in sockets, buy some spares and keep them on hand I guess.

Yep. Fortunately 1488s and 1489s are cheap and plentiful. And truth be known, they're actually extremely resilient...Visualize the harsh electrical environment of typically dozens to hundreds of feet of wire, running past things like blowers, lighting fixtures, etc, then being connected directly to an IC pin. Yes, they die, but only after the most egregious of insults.

Technically ethernet has magnetics, it's isolated. Most likely it's the transformer module that gets blown out. But I have never tried to fix one. The transformers keep the ground loops from happening I'd imagine.

For "DC" stuff like ground loops, sure, but when a pulse enters a transformer, a pulse exits a transformer...and if it's a big pulse, well.. -Dave -- Dave McGuire, AK4HZ New Kensington, PA

The Navy uses fiber where possible for the same reason, especially aboard ship. Lots of lightning and fun EMI concerns out there, plus a horrible thing in the MIL STD called "salt fog".

Makes sense! On a bigger ship where systems are in offices well away from the outside, do they still get corrosion from salt and stuff? - Ethan