The 'green' switch runs full speed and stone cold. The older ones ran hot.
Yeah, isolating networked faults can take time, if it's not obvious one goes into sledgehammer mode, verify each piece of the puzzle in turn until the fault is found.
Except when one has to find a bust cable, replacing each cable in turn until the fault cleared :( Only had it happen on one customer's site (last century).
Grant.
Yes, but I have different sets for Winter and Summer, and a few for wearing around the house (heavy during the Winter months) or weekends. I buy about a dozen pairs of each, when they're needed.
I've never seen "odd" sized slacks.
In the mid-'70s we had an AC breakdown in IBM, so they let us discard the ties (I ignored the fact that it was in a completely different building). After that, I never went back to wearing a tie.
I have a suit for weddings, funerals, and interviews, though the directions for the last few interviews have been "business casual". BTW, the suit is black. ;-)
The kitchen is hers. I'll clean the plates off after diner but that's about the extent of what I'm allowed to do in the kitchen. ;-)
I have four pairs, black sneakers for work, white sneakers for the rest of the week, and two pair of old white sneakers for mowing the lawn or working in the shop. Oh, and I do have some boots somewhere but I haven't worn them since I moved South. ;-)
I did that for five or ten years. I was living in Vermont so much of the year was cold so hiking boots worked well. They were about all I could find that fit my feet anyway.
No, they wouldn't cut it down here (AL-GA).
Did they put *numbers* on it? E.g., Spehro's switch and the one that failed/fixed here each claim ~5W (5W vs 5.6W -- I don't see
0.6W as a significant difference)No different than any other troubleshooting -- except you might have to do a fair bit of *walking* to get to the extremes of the "system".
I've only once encountered a bad cable (retaining tang snapped off and the cable had wiggled its way out -- just enough -- of the connector). Switches have been a "periodic" problem. More often, something went down, fsck(1) failed and a system didn't go multiuser, etc.
In my case, many devices are headless so troubleshooting can be a chore. I keep a little 12" PS2 keyboard and 7" LCD monitor just for those occasions! (I may be a slow learner -- but, I
*do* learn! :> )
Or a *loose connection* (even more frustrating -- replace *every* cable and it's still there... then, it *isn't*! :< )
A TDR comes in handy, here.
Still, 10Base2 made relocating devices (within a room) much easier than 1*BaseT! No need to chase down "a cable that's JUST three feet longer", etc.
Oh, OK. Weather is reasonably constant here: warm, hot or hotter. So, little change in sock quality! :-/
slacks
31", 33", 35", etc. Far less common than 30, 32, 34, 36, etc. I guess men are supposed to gain/lose weight in 2" increments? (I find a pair of slacks 1" larger than "my size" will *fall* off me. No hips?)they're
shirts)
I did a stint at an IBM plant and went to work "dressed to the nines" (I didn't work for IBM; just troubleshooting a "tester" we had built under contract to them). Worked in a glassed in/secure room. With a name tag (that their security folks took very seriously! It must be *displayed*, not just tucked into your pocket so it doesn't interfere with the multi-rack device you are troubleshooting!). Felt like I was in a store's "display window"! (Or, a fishbowl!).
Ick.
I reluctantly bought a grey suit some years back. The others are navy, light blue, black pinstripe (zoot-suit? :> ), brown herring bone, etc. Lots of suits. I just don't like *wearing* them! :>
I prefer to do most of the cooking. And *all* of the stuff that
*I* am eating (if she wants to make something that only *she* enjoys, that's fine. But, if *I* am going to eat it, then *I* want to make it! :> )And, I'm a much better baker, sweets (candy, ice cream), etc. I think you need a certain amount of patience to get those things right (e.g., a cheesecake is 5 hours in the kitchen).
I'd go crazy keeping track of them! My "sneakers"/walking shoes are black. But, I have two different styles. So, if I happen to see some black shoes, there's a 50-50 chance that they are the *wrong* black shoes! :< (I keep my pair of dress shoes hidden in the back of a closet since they see such infrequent use)
And the salt eats them up! I would wear "cowboy boots" and usually have to replace them in the Spring -- salt stains around the bottom. (sigh) That's one aspect of winter I *don't* miss!
Sounds like you solved your problem for the time being, and I really can't suggest any particularly "good" 8 (or 5) port units (I've had many different brands die, or eat ports, so...)
I have been pretty happy with 24 port units (ES4324) from Edge-Core, but they only have a bit over a year on them so far. I do have 17 of them running, but that's not the same as having them running for 5 years or more...
Nice bang-per-buck, fiber options on 4 ports/switch, multiple sorts of smart-switch software toys. But they have fans. And are probably obsolete by now...
-- Cats, coffee, chocolate...vices to live by Please don't feed the trolls. Killfile and ignore them so they will go away.
I've never had a single port go bad. Is this typically a *silicon* (or transformer?) failure? Or, someone getting careless plugging or unplugging a cable and mangling a connector? (i.e., the latter could conceivably be repaired)
Yikes!
Agreed. At ~9000 PoHrs/year, you wonder how folks can avoid the problems that plague the "consumer kit"? E.g., when I recap a device, I typically use caps rated at ~7000 hours. How do these folks work around that (besides keeping the product cool to begin with?)
The stuff that is being *delivered* to you is probably
*also* obsolete! :<IMO, the biggest downside with obsolescence is that if you decide you *are* satisfied with a particular product, it's probably "no longer available" (except in a surplus/used market) by the time you come to this realization and go looking for more! :-/
Usually blamed on lightning. Hard to be certain in all cases. I'm forcibly cheap (tiny budget) - I just mark the bad ports and downsize the switch. In my case the fiber port aspect of the switches I bought last year was a required upgrade, as I tried to replace all inter-building connections with fiber to get rid of lightning as a blame-able issue. They have lost no ports so far. I still have a couple of wires on minor connections, which are run through "ethernet surge suppressors" but glass is far superior to copper for those outside connections, IME. We've already had one event where all the surge suppressors on the copper fire alarm system were replaced, and the fiber network ignored it. After the second time that happens I'll point out (again) that I have spare fiber and the fire alarm can be configured to use it. The guy in charge of the fire alarm claimed to be interested at one point, and then he claimed it couldn't be done - since then I've talked to many people and vendors that do it all the time, so I think he just got into territory he didn't understand and turned that into "it can't be done."
I run a small campus network. I know people with hundreds, but they are also prone to the Cisco/Juniper end of the price spectrum which I cannot afford. See "forcibly cheap (tiny budget)." One thing I did try to do in concert with that upgrade was to remove as many "office switches" as possible to make the network "flatter" since I generally have all the ports needed for a building on the main switch for that building - needed more wires to certain rooms to make that happen, but I was able to stop using a bunch of (100 Mbit, mostly) "little" switches. And I'm not sorry to see them go.
Yup. OTOH, the upside is that you can get "not quite bleeding edge" stuff really cheaply sometimes, especially in that used market. My fiber optic modules (SFPs - small form plugabble) were 4GB fiber channel that had been obsoleted by a higher speed FC - but they turned out to work just fine for Gigabit ethernet, and I got them for under $6 each. Yet another reason I don't play Cisco - from what I understand, Cisco stuff only wants to talk to Cisco SPFs/GBICs and those have an automatic $300 or so premium over standards-compliant interchange equipment.
-- Cats, coffee, chocolate...vices to live by Please don't feed the trolls. Killfile and ignore them so they will go away.
Oh, OK. But, that would imply *exposure* to lightning?
Ah! ^^^^^^^^^^^ OK, so that's where you're seeing toasted ports! (?) Once I leave the premises, it's wireless. Mainly to avoid the hassle...
Does building/fire code allow this shared use? Does it impose other regulations on the availability/reliability of your network infrastructure ("Gee, the fire alarm wasn't working because Bob was transfering a large file...")
OK -------^^^^^^^^^^^^^^^^^^^^ That explains needing ~300 ports!
Agreed. The one "commercial" Cisco switch I have cost a pretty penny. I think ~$20 per port, IIRC. I'm now looking for PoE capabilities and figure the only affordable way of getting that is to look for *older*/obsolescent kit! Maybe something with a toasted power supply (figure a 24 port switch would use a ~400W power supply) that someone ran off to the tip instead of having repaired!
Understood. My brain still clings to the "5-4-3" rules of earlier technology so I tend to keep things flat, regardless.
Understood. In my case, it's *my* money so I'm not likely to "upgrade" after the initial purchase. So, I need to get *close* to the newest technology -- yet far enough away that the heat (price) isn;t too intense!
Yup. (see my "ethics of forced supply purchases" thread :>) You *sortof* don't mind -- if it's a capability that you truly *need* (or, from which you can demonstrate lots of benefit) and there is no practical alternative. *And*, the price isn't so incredibly "over the top" that it's sinful! If you can also have the potential for embracing a third-party's product "at a later date" (and a reduced price), then the future safe aspect can have extra appeal.
[OTOH, when the future is measured in *months*, it becomes a bit of a game :-/ ]
My 24 port non-PoE gigabit switches use 14-17 Watts (measured) with a rated Max consumption of 30. A PoE setup is automagically going to add
50-60 watts per port to that (at least in supply size). So that's 1200+W for 24 ports. That's quite a power supply.Depending on your actual needs, PoE injectors may make more sense - they certainly take away the price premium on the gigabit switch part of the system. And you don't need all the ports to be powered if you only have a few PoE devices.
Even though it's "my workplace's" tiny budget, the thinking is pretty similar. 10 Gig ethernet equipment was available when I put it in, and it wasn't even remotely on the table. But I wasn't going to settle for a lousy 100Mbit over fiber, either.
One thing I did do was spend a few more of my limited dollars on getting
17 switches all the same, even though some points would be served perfectly well by a smaller (thus slightly cheaper) switch. This way any switch can be moved to any other position in the event of a failure, and they all have the same management software/interface (which helps my sanity.)-- Cats, coffee, chocolate...vices to live by Please don't feed the trolls. Killfile and ignore them so they will go away.
No, typically they are much less! PoE only gives you ~12.5W (usable) per drop. Figure 15 (makes the math easier :> ) at the switch. That's ~360W above the needs of the switch.
Many PoE switches don't even supply the full ~15W per drop as, I think (?) many installations use them for PoE powered VoIP phones. So, you'll often see just ~7W per drop -- cutting power even more.
My objection (besides cost! :< ) to commercial PoE-enabled switches is that you power the whole d*mn switch, all the time! In a commercial setting, this is no big deal -- put a UPS on it (or a whopping big battery) and you're safe for most reasonable outages.
But, *here*, if the power fails, I really only want to make sure the port to the "outside world" (via "whatever" interface) stays up *and* one or two VoIP drops. I don't want to power any other nodes. Nor the portion of the switch that would *talk* to them!
This sort of capability is too "specialized" (apparently?) for current offerings. I think that will change as networks become more ubiquitous *in* SOHO applications as the user base "matures". (I just am not excited about *waiting* for that to happen!)
Yup. Though you still have the switch issue to contend with.
I can work around it by putting in a smaller switch that services those nodes that I *want* to keep powered and/or active. But, that means another piece of kit, *sizing* it for my needs now and hoping they don't change, etc.
It would just be so much easier if I could treat ports as individual units *in* the switch and decide how each will behave inthe absence of power, etc.
[I looked into designing a switch so I could have that sort of capability. It's *hard*! :> Obviously, the bandwidth requirements dictate specialized solutions, not "general" ones adapted to the application. :< ]Here, even the 100Mb links are more than adequate. I'm serving X terminals, SMB/NFS shares, basic network services, etc. The only time it really feels *painful* is when moving entire volumes, for example (I recently moved a 2T volume over the wire. I will NOT do that again! "SneakerNet or bust!" :> )
Yes!!! I tend to buy kit in pairs for the exact same reason. If something breaks, I don't want to have to mess around trying to get something *else* to fill in for it "temporarily". Easier to just pull "the other one" from wherever it happens to be and start using it in its place!
I was just commenting, last night, about what a nightmare it must be for IT folks to have to maintain different *hardware* platforms, let alone different versions of applications, different OS's, etc. I have a tough enough time managing the few different systems I run, here -- despite keeping verbose logs for each, etc.
NoThankYouVeryMuch! :>
Of course, I've never *tried* to learn more efficient ways of doing this. Rather, identify the big time syncs and figure out how to eliminate/reduce them! Not very sophisticated! :< It's also a huge part of why I don't update things often -- if it ain't (badly) broke, don't fix it!
(You have my sympathies)
you can join patch cables with the apropriate plug-in joiner, for a temporary solution it's often good,
That unit designed to fit a wall-plate, but can be used in-line too. They work ok on gigabit circuits.
-- ?? 100% natural --- Posted via news://freenews.netfront.net/ - Complaints to news@netfront.net
Hmm - I guess I've been polluted by too much exposure to folks trying to push the upper bounds of PoE capability, who are heading in the
50+W/port direction. The only thing I actually use with PoE came with it's own injector (and that's a proprietary scheme anyway) so I haven't looked into in in much detail, but I get newsvertising that has "so-and-so's scheme for 50W/port approved by XYZ" frequently.I certainly don't know of any that will kill ports in a power outage if the thing on the other end of the port doesn't die (well, short of "with an SNMP system telling them what to do under each condition - which might come close to what you want, but would require the SNMP system talking to the UPS to be up at least long enough to configure everything into "power outage mode" before taking itself offline to save power.) I also have yet to find an SNMP system that meets both "I can afford it" and "does what I think SNMP should be able to do for me", but in theory it could happen.
Anyway, most of the newer greenwashed/ecosmart/whatever switches actually do power-down (or power-reduce) ports not in use, especially for commercial applications, because wastage in commercial data centers has been an area of interest to address for LEED, etc. Some claim to even reduce power to shorter cables. Things multiply when each device needs power - size of UPS/Generator, size of cooling plant, power for cooling plant, etc - so the savings when you have a rack full of the stuff (and a building full of racks full...) is far more than the base power savings.
And then you start in with advertising how green your datacenter is...
You might look at:
Cisco 8PORT Sg 200-08P SLM2008PT-NA Gigabit Poe Smart Switch
Netgear GS108P ProSafe 8-Port Gigabit Switch with 4-Port PoE
Edgecore does make one in that line/size (ES4308-PoE), but it's more expensive than the above and out of stock at the least-expensive US dealer. They also make far more expensive ones.
The netgear "pro" stuff is somewhat better built than their plastic consumer stuff, and may have a lifetime warranty (they did at one point, I don't know if it's still true.) I suspect that cisco "small business" is a "linksys cisco" but it gets harder to tell as time goes on. Both claim green power management features that are pretty typical for modern switches. They are hardly going to be alone in that. I haven't used any of them though - just some of their predecessors/relatives, and both linksys and netgear stuff has gone south on me in the past. So have many other brands (trendnet, d-link, etc....)
Somewhere back there you asked about the fire alarm being held up by data transfer and I slipped past it - no, the fire alarm (if it came to fiber here) would get its very own fiber, or pair of fibers, from the 12 I have in my OSP cables. On the present setup I need 4 and occasionally use up to 8 since I have them for the base networking stuff. I'm also saving a couple for the phones so that they won't have to share with the computers if the phone system ever gets updated. If the cable got cut there would be a repair issue, but that's going to happen with either type of cable. If we were not so cheap we'd have cut-tolerant redundant cabling (the switches are smart enough to handle that), but we are too cheap for that, and the present copper fire alarm is not cut resistant in any way either (in most places the systems share conduits between buildings.)
I can also tell you that, having used it in the past, I'm delighted not to be fooling around with wireless between buildings anymore. At least in my setup, the distances required always made the links at least somewhat sketchy, and always slow. Heavy rain made them worse. Wireless to connect to laptops "so they are connected and can read email, browse lightly, etc." - sure. As a network link - hate it. To connect to anything including a laptop that needs to move serious data - find a wire. Between buildings - glass.
-- Cats, coffee, chocolate...vices to live by Please don't feed the trolls. Killfile and ignore them so they will go away.
Yes, I have something similar (though *intended* for in-line use) that is easily "broken apart". I did so with several of these, painted them red with a large 'X' and use them to make crossover cables out of standard patch cords.
I can *always* get a straight through cable that's "about the right length" (or make one). OTOH, I don't keep XOVER cables (out of fear I will get one mixed in with "regular" cables. The little (now bright red) "splices" are a convenient way to keep one on hand (think: laptop case) without incurring the risk/bulk of an actual cable!
I have these self-adhesive "clips" intended for anchoring cables within an enclosure (no idea how to describe *or* google for them). I staggered them along the underside of my work surfaces. Network cables come up to (from floor sitting machines) or down (from things sitting atop the work surfaces) and get channeled into these successive clips until they reach a switch.
[Of course, the {up,down}links from the switch also run through these clips to get to other switches]This keeps the cables off the floor (which is already littered with power cords, SCSI cables, etc.) and keeps them accessible (though not particularly convenient!) -- instead of hunting through
*bigger* cables (SCSI, power, etc.) that would also be present if the cables ran along the floor. ["Pulling" a network cable that doesn't have a boot in such a tangle is almost guaranteed to tear the little "tang" off! :< ]
PoE+ has higher capabilities. Here (residence) I am primarily interested in using it for VoIP phones (which, obviously, are a win *only* if powered from the same network cable! The idea of a phone having a cable AND a power cord -- wall wart -- seems counterproductive), STB's, comms radios, "network speakers" and various bits of automation. I figured ~10W per node was a modest budget... I'm not looking to power the *TV* off the network! :>
[The toughest power budget was the "network speakers" as you really *do* want to have some appreciable power to deliver to the loudspeaker without having to resort to a local power source!]My Cisco AP's have their own injector and power supply. I don't think they are technically compatible with the formal PoE spec but just "use spare conductors as power". (The actual spec is more precise in *how* you use those conductors!)
I wanted to embed this technology in the switch. I.e., configure it so that it knew what to power OFF in absence of mains power.
Ideally, I also wanted it to power down the individual ports and
*watch* for the PoE signature from the PD to determine when it *should* power up the device. Imagine having a VoIP phone that's idling on scant little power -- just enough to detect the handset going off-hook. Then, signalling the switch to power up the associated port (if so configured!) so the phone could be *used*.At the same time, the switch needs to be managed so it can be commanded to power up/down ports at will. E.g., you want to listen to music in the dining room? Then power up the network speakers, there, and route some music to them!
Yup -- as in my case of wanting to keep the UPS small...
For the PoE uses, I don't need that speed. I.e., phone, audio even video run well enough on 100Mb. Heck, even *10Mb* fabric would probably suffice (if it's *switched*!)
I don't think "name brand" is a reliable predictor of quality, anymore. I thikn you have to seriously consider the actual item, its intended market, etc.
Then, hope! :>
OK.
I've used the Cisco radios as a bridge between my house and neighbor. Safer/easier than cable. But, it was only a temporary setup.
I'm not fond of wireless at all since it lets "signal" go places where I might not want it! We've talked about using RF blocking paint inside next time we redecorate...
That would be silly. The PoE way of doing things is perfectly obvious and there really isn't any reason not to do it, at least, in a similar way. All of the hardware is already there. It's not "just use the spare conductors", rather injecting power/ground into the center-taps of the Tx and Rx transformers. There are often four transformers per package, so do the same to the other wires, too (essentially four wires in parallel for each the power and grounds).
In Cisco's defense, I think the AP's in question predated the formalization of the PoE spec (??).
AFAICT, they just push power up the two *unused* pairs and pull it off at the far end. And, I think their pre-PoE
*switches* had an even different (inband signalling?) scheme. [I think there are several other "kludge" products/vendors that use the unused pairs and ad hoc voltages, etc. -- PoE should be nominally 48VDC]PoE, by contrast, says you use the spare pairs *or* the "signal" pairs termed midspan (mode A) or endpoint (mode B), respectively. The PD has to be able to tolerate *either* mode A or mode B as it has no say in how the PSE will decide to deliver power! [Above 100Mb, you can only do endpoint because all four pairs carry signal]
For endpoints, you're passing power via the centertaps of the PSE to the PD. For midspan, there are no transformers involved :-/
But PoE tries to deal with conditions that it *expects* will be encountered. I.e., a crossover cable would cause polarity of the delivered power to be inverted so the PD has to expect that. A non-PD device could be plugged into the drop so the PSE needs to be able to identify "likely PD's" (consider telephones plugged in mistakenly -- with RJ11's *or* 45's!). So, the PD has to interact with the PSE -- and the PSE can't just blindly provide power on pairs (since it doesn't know beforehand that the device on the other end will be a PoE PD!
I.e., I have no hope that my AP's would satisfy all of those requirements -- given the way the current "injector" appears (little more than a place that connects a 48V wall wart and an incoming CAT5 cable to an *outgoing* CAT5 cable attached to the AP)
I can live with that. :>
The nice thing about re-capping the offending kit is it will now last a long time and won't crap out at an inopportune time. Our company makes a 5-port gigabit Ethernet switch for one of our products which hasn't had any problems. I'm glad we went that route since commercial switches we considered using generally crapped out within a year due to capacitor issues. I marvel how these guys can sell a switch for $30. Costs us a lot more!
Well, there's no guarantee that it won't crap out at an inopportune time! :> (Even if it did, you'd tend to only NOTICE that fact when you *needed* it! :-/ )
I think switches just have exhorbitant duty cycles. I.e., you
*never* turn them off.I've been tracking performance of eCaps in LCD inverters (another excellent case for them failing "frequently") and it seems like
7000 Hrs is where they crap out (based on several instances of a particular monitor that *tracks* "active hours" (i.e., total PoHrs might be 30000 but only "lit" for 7000 of those)With enough data points, I am hoping to see what *real* life ratings of caps are (vs. "specified" values)
Later versions of 'Belarc Advisor' display a list of all computers on your network, making the missing machines obvious. Then you drag out the keyboard & monitor. :)
You don't keep plenty of spare 1', 3', 5', 7', 16', 25' & 50' cables on hand? ;-)
I know that it's overkill, but they have an unused air conditioned server rack that they don't think they can sell, and... ;-)
-- You can't have a sense of humor, if you have no sense.
"ping machinename" (or, "ping IPaddress") works pretty well :>
But, that doesn't tell you if the current machine is misconfigured (wrong subnet, etc.), unplugged, broken switch, broken cable, or something in the targeted machine.
E.g., one of my file servers appears not to be booting. It won't even talk to the network *until* the OS is running in multiuser mode.
Unfortunately, it is *designed* to be headless -- no display
*or* keyboard. So, I have to talk to it via a serial port (run tip(1) or something similar on another machine over a serial cable to that server)Shortest cables are 3 feet and I don't think I've ever used one (I have a 1' cable that ties a NAS box to a wireless router located underneath it, but thats unusual). Most of my cables are 7-8', 12-15', 25' and > 50'.
Since I affix the cables to the underside of the work tables, if I move a piece of kit 3 ft, then I really need a cable that is *just* 3 ft longer than cable is currently being used to connect that device to the switch.
I finally learned to label the cables with actual lengths. Makes it a lot easier to find a cable *without* having to remove the existing cable and use it as a yardstick in picking a *new* cable for the job.
Wow! That's pretty cheap! I'll have to look at it to see what the "catch" might be (only has a total PoE budget of 3W :> )
Exactly. I am trying to size the fabric to fit the actual needs.
I've thought of getting four ~40U racks and setting the guest bed atop them. This would give me a place for the various bits of rack-mounted kit that I have (freq synths, progr power supplies, logic analyzers, etc.) without taking up more "floor space".
[I am *really* intent on getting a BASEMENT -- even if it means "elevating things" to be able to get under them! :> ]I suspect the elevated guest bed would meet with some "resistance"
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.