I just bunged a gigabit switch on my LAN. Blimey, copying NAS1 to NAS2 via a pc (windows drag and drop), the pc is peaking at ~300Mbps. On the 100meg LAN it managed ~60Mbps. I ought to try FXP between the NAS's though I spose the drive write speed will make it similar. How the devil do they get gigabit to work? What kind of signal level do we get, a few milli-volts? I remember donkeys years back, if you hung a scope probe onto a 1meg line it'd kill it dead.
For many purposes, gigabit is in the category of things that are, quite simply, fast enough. And you often wouldn't appreciate any further speed up. Pretty much where we are in terms of processors which hang around most of their lives waiting for something to do!
Obviously, people can come up with things that would benefit from more of any performance factor, but I don't think many domestic installations need more speed sufficiently that they'd actually pay for it.
Gigabit is a good match to the real throughput you're likely to get from a 7200 RPM disk. I can get just about gigabit throughput backing up between my systems with 7200 RPM disks (although I had to do some tuning of buffering to get there).
IIRC, it's the order of a volt. It works by sending the data over all four pairs in both directions at the same time. The actual baud rate on any one pair is the same as
100Mbit ethernet (which was already at the 125Mbaud limit of Cat5e cable), but the use of all 4 pairs and additional voltage levels to encode more bits/baud enables it to achieve 10 times the throughput.
I recall putting a scope on a 10Mbit coax ethernet. If you triggered on the start of a frame, and did an ftp file transfer, you could see things like the high order TCP sequence number bits counting up in binary (if you're used to watching an incrementing binary bit pattern). One thing that surprised me was how stunningly different the signal amplitudes were from different systems on the ethernet - if there were only a few active systems, you could use this to identify them.
In message , at 19:27:27 on Mon, 18 Mar 2013, polygonum remarked:
Pretty much agree. 300MB file *to* a NAS is about a minute, *from* it twice as fast. Although I bought an awful NAS last year (didn't believe the reviews) which really does take several days to absorb 500GB.
I'm on my second generation GB hub now, the first one from about five years ago took so much power it almost glowed in the dark. The new one takes a tenth of the power, and also has 8 ports instead of 4. Decent tower PCs of that vintage have GB on the motherboard.
Anyone want a glowing 4-port hub. Cheap? I thought not.
Got gigabit around six years ago - almost by accident. Needed new router and it happened to come with gigabit. Then got machines that could use it. :-)
I can't answer your questions but I was also Dead Impressed when I first gave the gigabit switch some real work to do. I immediately laid plans to rewire the parts of the network that were sharing LAN and phone on the same CAT5 cable.
It's not often that the new technology works ten times as fast as the old one.
Yup for bulk transfers it makes a big difference...
I was at a clients other other week migrating someone to a new PC, IO noted he had 12GB in his email folder, and had the sinking "I am going to be here a while" feeling. Then I remembered I had replaced their switch with a gigabit one a few years back. Brought the whole thing down to under 30 mins ;-)
How does that compare with sticking the old computer's HDD in a USB enclosure, plugging it into the new one, & copying from there? (This assumes, of course, that the old one is being decommissioned.)
Well in part that would depend on how far away they were on the coax. Remember there was DC resistance that lowered the signal amplitude with distance and thereby limited the overall cable distance (and also defined the minimum allowed frame length). Collision detection relied on two colliding signals raising the DC level above some value.
This applied to thinnet too, but didn't stop some physicists at SLAC thinking they knew better. One segment was already at the maximum length, but some smart-alec extended it with some of their 2mm thickness
50 ohm cabling they used in their equipment. Course the collision rate went stratospheric and throughput tended to zero. The Ops guys gave them a bollocking over that.
Those guys also had a habit of disconnecting the thinnet T from the back of their workstation, and inserting a long tap cable from there, no the cabling looked neater. That halved the impedance, with similar effects on performance.
10mb/s >>> 100mb/s They made a big thing about "fast" ethernet being backward compatable: "10/100Mb/s" but they don't do that with Gigabit, even though it will fall back to 100Mb/s and I think 10Mb/s too.
Its comparable - with gig ethernet you can cope with the full rate of most normal hard drives. (having said that I usually take my USB to SATA/IDE interface to such jobs anyway since its often easier to just hook the drive up directly.
Still has to be isolated even with PoE, except for a special case of a terminal device which is in an insulated plastic case with no connections other than the single ethernet RJ45 itself (such as a wireless access point with no external aerial connection).
There were some holes in the spec which didn't cope with having just a 2-pair cable when doing gigagbit negotiation, although I think all chipsets now spot that failure and renegotiate at
100Mbit max when it happens.
Again, some early 10GBASE-T interfaces were 10G-only, and couldn't negotiate down.
HomeOwnersHub 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.