Sixty times faster than a T1!

[SageBrush]

Originally Posted by daniel

Comcast tells me I have 12 Mbps download speed. I don't remember what they say my upload speed is supposed to be. Speedtest.net tells me I actually have 28 down and 8 up. Better than double what Comcast promises me.

I'm skeptical. Try another speed test, and be sure of the measurement units.

[RobH]

Originally Posted by jayman

Billions of taxpayer dollars were shoveled into developing the system, and at first the hardware was exotically expensive. For example, the first Interface Message Processor was built by BBN (Bolt, Beranek, and Newman) and used a military version of the Honeywell DDP-516. The original cost was millions per unit

I worked with a programming group that used host # 8 of the Arpanet. Network addresses were 8 bits, with no name lookup service. Actually, 8 bits overstates it, as it was actually 6 bits of IMP address, and the remaining 2 bits the host number on that IMP. The network backbone ran at the incredible speed of 64Kb.

TIPs (an IMP that connected to computer terminals) supported about 10 users, typically at 0.110Kb. The first upgrade was to 0.300 Kb, and I think they even went to 1.2 Kb (not sure about that one - maybe it was just the IBM computer that was upgraded to 1.2 Kb per terminal). One of the arguments against squandering bandwidth of 1.2 KB on a terminal was that no one could read that fast.

We had both the ruggedized IMP and an ordinary DDP-516. As I recall, the ruggedized version had its core memory set in epoxy to help survive a bomb blast. The initial testing of the core memory set in epoxy didn't work, as the epoxy held the core memories so tight that the bits couldn't change. The solution was to cycle the memory as the epoxy was setting. That left enough wiggle room for the cores to flip state.

Then there were the 3" diameter eyehooks on the top (and bottom?) of the IMP. They were there so that the IMP could be mounted on springs to help survive that bomb. Ours just sat on the floor like the rest of the equipment, but it still had the eyehooks.

Our IMP was connected to an IBM computer that was used part time in secure computing. When secret data was being used on the computer, anyone not on the approved list was prohibited from crossing the white line marking the perimeter of the computer. It got to be such a hassle that the IMP was finally moved to a non-secure lab on the next floor. It made the security types so much happier that the geeks were finally excluded from access to the secure computer. Apparently nobody ever explained to them that that computer was connected to the IMP, and thus to the rest of the Arpanet.

[apriusfan]

Originally Posted by Dan-Wolfe

T-1, also known as DS-1, I believe has a speed of 1.544 Mbps. It was originally designed to carry 24 voice calls. This standard has been around for ages.

I suspect that T1 appears faster because unlike cable and DSL, it's not shared.

How is DSL 'shared'?

[daniel]

Originally Posted by SageBrush

I'm skeptical. Try another speed test, and be sure of the measurement units.

Interesting!

Speedtest.net connected to Liberty Lake (a stone's throw from here: I can drive my little Xebra there and back several times without recharging if I want to) gives 29.69 Mbps and 8.48 Mbps; connected to Seattle it gives 31 and 3.61. A friend of mine nearby gets similar results.

Speakeasy.net gives 3125 Kbps and 6766 Kbps respectively the first time I tried, but 15 minutes later it gave 30515 Kbps and 2331 Kbps.

Auditmypc claims to have the "most accurate" test, and offers two different kinds, and I got two very different results: 14.75 and 10.19 Mbps for one test, and 3672 and 4375 Kbps on the other test, which is pretty extreme difference for two versions of "the most accurate" speed test.

I recently switched from a very old modem to a new one, and I can tell that my speeds are much faster than they were.

[a_gray_prius]

Originally Posted by apriusfan

How is DSL 'shared'?

There are a *lot* of DSL lines hooked up to the CO (central office) DSLAM (DSL access multiplier). So there might be 500 1Mb/sec lines connected to a 10Mbps pipe.

[apriusfan]

Originally Posted by daniel

Interesting!

Speedtest.net connected to Liberty Lake (a stone's throw from here: I can drive my little Xebra there and back several times without recharging if I want to) gives 29.69 Mbps and 8.48 Mbps; connected to Seattle it gives 31 and 3.61. A friend of mine nearby gets similar results.

Speakeasy.net gives 3125 Kbps and 6766 Kbps respectively the first time I tried, but 15 minutes later it gave 30515 Kbps and 2331 Kbps.

Auditmypc claims to have the "most accurate" test, and offers two different kinds, and I got two very different results: 14.75 and 10.19 Mbps for one test, and 3672 and 4375 Kbps on the other test, which is pretty extreme difference for two versions of "the most accurate" speed test.

I recently switched from a very old modem to a new one, and I can tell that my speeds are much faster than they were.

You may be seeing the impact of sharing of the cable connection with the different speed results. An objective network bandwidth availability analysis would cost $ that could be better spent elsewhere.

[apriusfan]

Originally Posted by a_gray_prius

There are a *lot* of DSL lines hooked up to the CO (central office) DSLAM (DSL access multiplier). So there might be 500 1Mb/sec lines connected to a 10Mbps pipe.

According to that definition, a T-1 connection is 'shared' as well. The typical definition of shared vs. dedicated is from the Central Office to the end user termination point. Cable bandwidth is truly shared between all of the users on the sublink. Get a group that are down or uploading files (like Youtube videos) at the same time and the response time goes into the toilet. With DSL and T-n, the only way your response time drops off (barring line failure between the CO and the termination point) is if the connection between the Central Office and the rest of the web tanks.

[jayman]

Originally Posted by daniel

Better than double what Comcast promises me.

I've had cable service in other areas, with rapid growth, and initially the speed was quite good. But they couldn't keep up with rapid growth and subscriber signups, and then the speed and connection quality really tanked

Originally Posted by RobH

We had both the ruggedized IMP and an ordinary DDP-516. As I recall, the ruggedized version had its core memory set in epoxy to help survive a bomb blast. The initial testing of the core memory set in epoxy didn't work, as the epoxy held the core memories so tight that the bits couldn't change. The solution was to cycle the memory as the epoxy was setting. That left enough wiggle room for the cores to flip state.

Then there were the 3" diameter eyehooks on the top (and bottom?) of the IMP. They were there so that the IMP could be mounted on springs to help survive that bomb. Ours just sat on the floor like the rest of the equipment, but it still had the eyehooks.

Correct, the military version of the DDP516 was designed to handle extreme shock loading, eg a nearby bomb blast. I now highly doubt it would have resisted the EMP and the entire network would have latched up

If I recall, the eyebolts were for lifting the thing, and the skid at the base had slots for a forklift truck or a handtruck. I've seen a couple mounted on springs, but the usual approach by the late 1970's was to have a "floating" room for the equipment

It was easier and more reliable to design the floating room, as it offered consistent performance

Originally Posted by daniel

I recently switched from a very old modem to a new one, and I can tell that my speeds are much faster than they were.

I had an old Terayon cable modem. When Shaw upgraded me to a newer Motorola SB5102, my download speeds easily tripled. Motorola claims this cable modem supports 30 Mbps download speeds, and is DOCSIS 2 compliant

Originally Posted by a_gray_prius

There are a *lot* of DSL lines hooked up to the CO (central office) DSLAM (DSL access multiplier). So there might be 500 1Mb/sec lines connected to a 10Mbps pipe.

Correct. Typically the line drawers BIC/DBIC's have to be properly engineered to account for customer DSL use. You can use add-on DSL, the so-called "pizza box" upgrade, but the quality is far more iffy

Originally Posted by apriusfan

With DSL and T-n, the only way your response time drops off (barring line failure between the CO and the termination point) is if the connection between the Central Office and the rest of the web tanks.

True that DSL is more consistent but spikes in Internet use can still cause problems, especially if poor attention is paid to how DSL subscribers are assigned in line drawers

[hobbit]

Back when I was doing a little telecom/pbx maint, we had a couple
of T1s hung off the main pbx -- one with a microwave shot to
another office, and one upstream to the telco. It was fun to
hang a scope off it, get the timing stable, and actually see
the 24 little slots in the whole frame. I could tell which
channels were active or not, and in many cases could "see" people
talking as the little block of bits changed. And even seize a
particular trunk for an outgoing call and watch my own "test! test!"
bursts show up.
.
_H*

[jayman]

Of course, you never ever purposely busied out any particular channels to annoy people

I am a bad man