It's the slew rate that's bugging me. How the hell can you get a decent sized signal at that speed? The slew rate must be off the scale or there's something (multitudes) that I don't know. We're talking nanoseconds here. A gigabit = 1 ns, yea? That puts RTL, TTL & CMOS out of the picture, shirley? Bollocks to it, i'll collect my pension as normal this week :D My claim to fame - I remember 75 baud or was it 110? teletype 33's :)
Are you quoting the transfer times for the "awful NAS"? I did a double take at your time of one minute for 300MB and tried it myself.
For a 366MB file I get:
6.0s [61MB/s] to a QNAP NAS (SATA RAID 5) 7.7s [47.5MB/s] to a FireWire 800 drive (SATA) 9.8s [37.5MB/s] to a homebrew NAS (IDE)Source drive in all cases a SATA I 7200 rpm drive. This may be limiting the transfer rate because the Blackmagic disk speed test gives 123, 62 and 110 MB/s respectively. Yes that is the correct order ,tested with a 1GB test file.
The homebrew NAS was running a backup job at the time which may explain why it was slower.
The figure you are quoting, 5MB/s is more typical of a transfer over
100MB/s Ethernet.
Do you mean Mbps or MBps?
Your figures look wrong in either unit. Gigabit Ethernet is capable of
600-800 Mbps sustained transfer speeds. So your figure looks about half of what one would expect or if you mean MBps that would be almost three times the expected transfer rate. OTOH presumably you are talking SMB on the NAS and Windows copy is hardly efficient either. But it seems wrong to only be getting half the practical transfer rate.
It doesn't need to go that fast. As Andrew Gabriel explained earlier, the transitions only take place at 125MHz. What gives it the extra speed is using all four pairs in both directions at once combined with having multiple voltage levels. There is equalisation in the transceivers to help compensate for frequency dependent cable losses and to cancel the echo of locally transmitted data at each end.
Thinking about the cable being "voltage" or "current" driven isn't really helpful. It is more useful to think of it as a set of balanced transmission lines carrying radio signals. The cable is driven from a balanced 100 Ohm (approximately) source impedance and it is terminated with 100 Ohms to avoid reflections.
John
there is something you don't know. Trellis code modulation. That squeezes te last drop out of a transmission channel
Data rate of a line is the bandwidth of the wire pair times the signal to noise ratio times the number of cables.
in gigabyte you have 4 pairs each capable of at least 100MHz and with (over a short distance) at least 60dB S/N.
I think they encode at 5 levels in that 60dB range so that's a top rate of 5 x 100M x 4 = 2 GIGABITS a second total. 1 GIG up one GIG down..
Nope. what's on the wires is analogue. By the time its digital its 32 or
64 bits wide. divide a gig by 32 and its only 34MHZ clock rate sort of speeds.If you like the secret of throughput is to do a lot in parallel
And use every corner of spectrum
indeed. current loop telex machines. ahah
On Tuesday 19 March 2013 01:45 brass monkey wrote in uk.d-i-y:
Try 10gig over copper. It exists and I have been told that the signal is rather less than the noise - there is much cleverness involved...
Was that developed in North Wales?
Yeah and the maximum range spans 1 metre to 15 metres depending on the interconnect chosen. With an amusing price range of £500 to £1500 for a PCI card.
In message , at 02:03:33 on Tue, 19 Mar 2013, Steve Firth remarked:
No, that was the "good NAS", but it's not a high performance one.
Trying again this morning, on a PC with Win7 (so it has a Bytes-per-sec meter) I got 15.3 MBytes/sec up and 29.1 MBytes/sec down. (20sec and 10 sec for a 300MB file). That's 125 Mbits per sec and 250 Mbits per second.
The "awful NAS" manages about 4MBytes/sec, which is 14GB/hr, or 35 hours to fill the entire 500GB. I suppose how many days that is depends on whether you count 8hrs/day or 24.
Well, its taken ~17hrs to shift 800gigabytes. The peak I saw was ~300megabits/sec so NAS to pc to NAS is about right for 17hrs. I was just trying FXP but have to go out :)
The digital interface is generally 4 or 8 bits at 125MHz between the MAC and the PHY. The 4 bit interface using both edges of the clock.
MBQ
Don't confuse the data rate with the baud rate (i.e. signalling rate)... gigabit ethernet uses the same signal rate as 100Mb ethernet. Only on 4 pairs at a time instead of 1, and with a higher number of data bits per symbol.
I remember confusing the hell out of serial card vendors in around year
2k looking for cards that would support 75 bps... they were all most unprepared for the "how slow can it go?" question. (needed a synch serial interface for talking to modems designed to run over HF radio where the top data rate was 2400!
Similarly don't confuse the bit rate on the line which is higher than the nominal bit rate due to the encoding scheme used.
MBQ
Up to 100m with Cat 7 IIRC
Pricing is scary though. We use multiple 10GB links but over fibre (or internal to blade chassis). Vmotion is nice over that sort of bandwidth :-)
Some of the kit is beginning to appear on ebay now - not cheap for the cards and of course, the router that plugs into isn't cheap :-)
Darren
On Tuesday 19 March 2013 13:48 D.M.Chapman wrote in uk.d-i-y:
You lucky tart!
ESXi 4.1 cannot make use of bonded links so all mine are stuck at 1 gig. I might upgrade to 5.whatever if I get some balls enlargment surgery done.
Hard drives already max out gigabit connections. In fact I think SSDs are already well passed the Sata1 3Gb/s mark.
If you want to copy large files from one computer to another it is nice to have it happen as quickly as possible. Given the size of HD video files I think it would be nice to have a quicker network connection.
I don't care if hardware spends most of its time hanging around doing nothing, when I want something done I prefer it to be done fast.
:-D
yeah, it's a bit scary. We are on 5 now, looking to go to 5.1 although vcenter is all web based and rather different.
Darren
I can get a few times that over wood, by carrying a 1TB drive across the room... ;-)
The cable to my lad's room only has the 1-2 & 3-6 pairs connected when plugged in to the gigabit switch a PC is fine at 100Mb but the NIC in his I-MAC will not negotiate a connection unless I plug it into one of the router's 10/100 ports.
I have also seen an odd thing at a couple of sites at work.
Existing PC on what should be a Gigabit LAN only working at 100 New PC indicates it is connected at 10Mb/s but there is no meaningful connection in practice. Network tester indicates no dc fault on all 6 pairs and a visual check at both sockets show their are no split pairs. Length of cable, about 15 meters.
There is unlikely to have been any joints along the way although I could not absolutely confirm that.
As I was not ultimately reasonable for the cabling I escalated the fault to someone who was, but it was difficult to explain why the old equipment worked reasonably well but the new stuff wouldn't work at all.
If shifting lots of data any distance, never under estimate the bandwidth of the postal service...
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.