Domestic Cat 5 cabling and management

I've done it for our cheapo home LAN and had no problems - all cat5 solid cabling goes direct to loft switch without a patch panel on the way. Wouldn't risk it if there was to be any movement though.

Beacause you live in a faraday cage.

Because phones plugged into wall sockets don't get lost,stolen, or run out of charge.

Because you need a landline for ADSL so why not use it?

I would almost put money on it not working particulary reliably here. But hey if you can put up with phone calls with dropouts, distortion and having to continually repeat things or get things repeated that's up to you.

Because mobile coverage is anything but reliable or universal, the voice quality is poor and the delay annoying.

Because it's useful to have a number that will reach whoever happens to be home at the time.

Thats the advanced US plug, with an earth pin! ;-)

Adam

The message from The Natural Philosopher contains these words:

I think that's just it. All too often people don't think logically.

The message from The Natural Philosopher contains these words:

Clearly your lifestyle has certain similarities to mine.

The message from Pete Verdon contains these words:

Because if you were like most people, you'd find that many people phone on landlines -- though that's not a reason for you at the moment

Because you could phone out for "nothing" i.e. at no marginal cost on any reasonable landline deal whereas mobile deals don't generally have unlimited calls during the day

The message from The Natural Philosopher contains these words:

or simply experence permanent battery death remarkably quickly

The message from "Dave Liquorice" contains these words:

And because handsfree is generally simpler and better with a good office phone

IME, for unmanaged switches:

Occasionally you can get a bargin from eBay second hand (I got a Netgear GS108 8 port 1GB switch for £30; this make would normally be much more), but more usually I see second hand ethernet switches going for more much more on eBay than they cost brand new (if you search around).

I wouldn't bother with cat 6 for home use. My home network is 1Gb, which is pretty well matched to the data rate I can get from a pair of 7200RPM mirrored SATA drives. (Drives are about 75Mbyte/sec sustained, and mirroring gets me to something over 100Mbytes/sec sustained read rate, which is similar speed to 1Gbit network. To make good use of 10Gb, you'd need a filesystem striped across many disks, even when using Enterprise class 15k RPM disks, or to wait for solid state disks to replace spinning rust.)

I actually meant to include a comment there about 100M being somewhat short sighted given that many (most?) machines now come with gigabit as standard. The price differences you quoted (and which I have now checked on my usual sources) are greater than I thought - looks like the 10/100 stuff has slipped down in price and the gigabit gone up a bit.

I doubt many of us have the level of harsh environment I mentioned in our houses - so, yes, unlikely to use Cat 6 at home.

Moore's Law (doubling every 18 months) does apply to disk transfer rates. So Gigabit is fine now, and will be for 18 months (one doubling) and from then on will be a problem for certain uses. It's the *cable* I'd suggest forking out for, not the electronics - the cable is there for the long term.

Unfortunately not. Disk capacities don't lag too far 18 month doubling now, but all other disk performance parameters lag way behind this, which is why disk performance has steadily become a more and more limiting factor in performance of many computing applications. A quick calculation shows disk transfer rates have doubled about every 45 months over the last 25 years. That's one reason applications often require complex disk array infrastructures, to make up for the serious [relative] lag of disk performance behind the progress made in CPU performance.

Out of interest, here's a table of relative performance changes I use in a presentation I give on filesystem performance from time to time...

Rotational speed 3,600 15,000 4x I/O's per sec 30 300 10x Transfer rates 1 MB/s 100 MB/s 100x Capacity 150 MB 1.5 TB 10,000x

CPU performance 4 MIPS 400,000 MIPS 100,000x

The CPU performance improvement is pretty much spot on for doubling every 18 months. All the disk performance parameters lag behind, mostly _way_ behind. The exponential increase in disk capacity is not linear over the period, being faster in more recent years, which is why it's nearly on doubling every 18 months now. When SSD's become mainstream, there will be a giant discontinuity in disk performance. They're still too expensive and too small to use other than in some specialist situations, but that's changing fast.

(Note that Moore's Law is really a doubling of transistor counts on a chip every 2 years, although it's often used to refer to any exponential gain scheme.)

Maybe the new faster flash memory devices will be able to provide the performance jumps required to feed 10Gb? :-)

I stand corrected. That suggests gigabit will last much longer than I thought.

Unlikely. The big gain with SSDs (flash or not) isn't the transfer rate, it's the latency. Or to translate that:

The peak rate isn't much different. You get similar rates - which isn't too surprising, as they are usually hung off the same bus that was designed to handle disc traffic.

But when you ask for a sector, you get it immediately. There's none of this move the head to the right place, wait for the disc to spin to the right place, read some data you get with discs. It just so happens I ran a benchmark on a USB thumb drive the other day. While transfer rate is nothing special, latency (which is typically several mS for a disc) came out as 900 microseconds.

Andy

SSD speeds have improved quite a lot - 220 MB/s now available (or at least announced) from Samsung. And I would guess they likely have further improvements not far behind.

Added back a second one for comment further down...

Yes, indeed. That translates into the number of I/O's per second (IOPS). There are now read-biased Enterprise SSD's giving >50,000 IOPS (reads), verses 300 IOPS you get from todays Enterprise disks (SAS or fibre channel). This can give you an astonishing speedup of applications which use synchronous disk i/o, as they are predominately limited by the number of IOPS they can get through the disk subsystem. 80,000 IOPS is probably about the limit of what you can get across current disk interconnects in any case, and actually way exceeds what most of todays disk controllers can handle, given they weren't designed for disks that fast.

Yes. An Enterprise SSD should get something even better, although I don't have a figure to hand.

Rod coughed up some electrons that declared:

Generally, flash has fairly slow transfers in bytes/sec - just about matching disks for the best and most expensive flash.

However, flash does have the advantage of very short random seek times.

Cheers

Tim