http://www.wired.com/gadgets/miscellaneous/magazine/17-10/ts_burningquestion
-Hank
Hank Nussbacher wrote:
Burning Question: Does Internet Speed Vary by Season? | WIRED
It's an interesting theory, that temperature affects overall throughput. Their assumptions on other conditions that affect bandwidth consumption are off IMHO. Our own data directly refutes what Wired reported in this article. Summertime is our most heavily utilized months on our network on average. For SPs heavily laden with residential subs I think this is probably the norm. Then school starts and you have a pronounced drop in traffic (that includes a major dip when college begins and again when primary school begins). The rates slowly increase back to their summer time highs until the holiday season begins where they either remain steady or taper off slightly. The theory here is that the high-bandwidth users are too busy with holiday affairs to play games, download music/porn, etc. That is until after X-mas when consumption suddenly spikes in a very pronounced way (new computers for X-mas). This also corresponds to our biggest month for new service turnups and speed increases in our bundles. Late winter varies from fairly constant to slight growth. Our single biggest days are the ones proceeding a major winter storm, or if the storm doesn't cut power to large swaths of our service area then the days in the middle of the winter storms come out on top. Spring growth depends on the weather. Good weather means less consumption for us. Bad weather means more consumption. Our least busy month is May when the kids are the most busy. June and July again show a major turn around.
Bandwidth consumption is directly tied to your user demographics. If your SP is primarily business circuit then your traffic patterns will vary wildly from that of a SP with primarily residential circuits. Every SP is a little bit different. That's why some SPs set personal records for bandwidth consumption when Michael Jackson's memorial service was broadcast (including SPs less than an hour away from me) and other SPs (mine for example) didn't have a single user stream the broadcast and otherwise had a normal bandwidth day. Other than Wired making an assumption that all SPs have nearly identical traffic patterns, the article is otherwise ok.
Justin
I'm not sure the effects are so big compared to the actual speed that they are noticable for the average user. We also don't have any proper data available but we do (operating in NL) notice from time to time that periods with loads of rain can have influence on the stabillity and speed of DSL lines especially in older areas of towns where they still have paper/lead covered cabling instead of more modern PVC isolation. This as more visible when everybody still used 56k dialup.This may as well be a very local effect, the western part of our country is largely at or even below sealevel and very wet already.
However as these effects might get you a few kilobits extra from time to time that effect is not visible in overall usage statisctics, as soon as the sun comes out we see traffic levels drop to only rise again near september when everybody is back to school and the office. As far as traffic levels go, it's the rainy winter nights which make it into the recordbooks.
Grtx,
Marco
:-> "Hank" == Hank Nussbacher <hank@efes.iucc.ac.il> writes:
> http://www.wired.com/gadgets/miscellaneous/magazine/17-10/ts_burningquestion
> -Hank
There are TXCOs and OXCOs inside equipment for a reason. And rubidium
lamps as well, sometimes.
Seasonal variations in usage from the end customers are a fact of
life, instead. If your net is large enough you can even spot the
different habits about vacations, holidays and whatnot across the
different regions.
Pf
I read the article and the follow up posts and I wonder if we are all using the same definition for "speed" here. The article seems to imply you don't get 6 Mbps on your DSL line in summer because the copper is hotter and it's harder to push electrons down the link. That is clearly BS, the clock is ticking six million times per second, period.
Then it talks about traffic, which is very different than speed, at least in my book. If the intertubes are congested, you might get less throughput, but your "speed" is the same. And congestion cannot affect speed. That laser is blinking at 10 billion times per second whether the queue behind the port is full or not. (And don't tell me the laser is quiescent when the queue is empty, you know what I mean.)
So what are are talking here? Speed, throughput, congestion, packet loss, latency ... ?
Oh, and while I am certain it is true different networks see different peaks & valleys for different seasons & times of day, the "Internet" (whatever the hell that is) definitely has less traffic in summer than fall.
Yeah I have to agree...that article was poorly written and clearly the
author has little understanding of physics. "Velocity" has absolutely
nothing to do with Wiedemann-Franz law, we are talking resistance and
that determines current through the wire; not how fast the electrons are
flowing through the wire. Plus unless you are selling services on
Mercury or Pluto, the temperature extremes between summer and winter are
not enough to affect the resistance level of a copper in such a manner
that it would affect the current on the circuit.
We all have traffic utilization graphs, and, at least in my case, its
pretty obviously that trends are affected by seasons only in terms of
vacation, holidays, and eve if it's a Monday or a Friday. If we need to
do major backbone upgrades, we try to schedule around a holiday weekend,
because that's when we find Internet traffic to be the least (well from
a business utilization point of view).
Bret
I read the article and the follow up posts and I wonder if we are all
using the same definition for "speed" here. The article seems to
imply you don't get 6 Mbps on your DSL line in summer because the
copper is hotter and it's harder to push electrons down the link.
That is clearly BS, the clock is ticking six million times per second,
period.
Are you trying to say that the *actual* DSL speed, as synchronised between the modems at either end, is neither a) affected by the physical characteristics of the copper pair, nor b) variable?
I agree the article is woolly between line-speed, throughput, goodput, congestion, etc, but to say that DSL line speed is in any way fixed in the same way that Ethernet or PDH / SDH lines are is contrary to every DSL platform I've worked with.
(Also, 6Mb/s DSL doesn't equate to 6 million ticks per second in anything relating to pushing electrons onto the wire. Remember, it's modem technology, just faster - your baud rate is still much lower than your bps.)
Regards,
Tim.
I read the article and the follow up posts and I wonder if we are all using
the same definition for "speed" here. The article seems to imply you don't
get 6 Mbps on your DSL line in summer because the copper is hotter and it's
harder to push electrons down the link. That is clearly BS, the clock is
ticking six million times per second, period.
So you're saying that if I put in an 8Mbps ADSL1 connection, then I'm going
to get a guaranteed 8Mbps point-to-point back to the exchange, regardless of
the quality of my phone line, or the distance from the exchange?
That laser is blinking at 10 billion times per second whether the queue
behind the port is full or not. (And don't tell me the laser is quiescent
when the queue is empty, you know what I mean.)
Laser? Perhaps this is a different type of ADSL than most people here are
used to?
(I'm not saying that the article is right, but...)
Scott
I read the article and the follow up posts and I wonder if we are all using the same definition for "speed" here. The article seems to imply you don't get 6 Mbps on your DSL line in summer because the copper is hotter and it's harder to push electrons down the link. That is clearly BS, the clock is ticking six million times per second, period.
So you're saying that if I put in an 8Mbps ADSL1 connection, then I'm going to get a guaranteed 8Mbps point-to-point back to the exchange, regardless of the quality of my phone line, or the distance from the exchange?
Yes, everyone, I was imprecise. Please tell me all about baud and variability and such. Because that was the point I was trying to make.
That laser is blinking at 10 billion times per second whether the queue behind the port is full or not. (And don't tell me the laser is quiescent when the queue is empty, you know what I mean.)
Laser? Perhaps this is a different type of ADSL than most people here are used to?
(I'm not saying that the article is right, but...)
I admit I totally spaced on the fact DSL != ethernet when I was typing the first paragraph. But when I wrote the above, I actually thought to myself: "I better mention I'm talking about a 10G backbone link... nah, everyone on NANOG is smart enough to figure out what I meant."
End of day, the point stands that the article is worse than useless as it does not add data to the general knowledge pool, but actually makes everyone dumber for reading it. Apparently it even made me forget how DSL works....
Scott Howard wrote:
<snip>
So you're saying that if I put in an 8Mbps ADSL1 connection, then I'm going
to get a guaranteed 8Mbps point-to-point back to the exchange, regardless of
the quality of my phone line, or the distance from the exchange?
<snip>
(I'm not saying that the article is right, but...)
ADSL systems will retrain to a lower rate as line conditions (SNR)
change for the worse. The attentuation characteristics of a given pair
will change of time due to a number of factor, including but not
certainly limited to physical wear, moisture invasion, localized source
of interference, sunspot activity etc.
Are dsl plants subject to localized environmental conditions? Absolutely.
joel jaeggli expunged (joelja@bogus.com):
ADSL systems will retrain to a lower rate as line conditions (SNR)
change for the worse. The attentuation characteristics of a given pair
will change of time due to a number of factor, including but not
certainly limited to physical wear, moisture invasion, localized source
of interference, sunspot activity etc.Are dsl plants subject to localized environmental conditions? Absolutely.
I could be convinced that extremem heat could change a cables velocity factor, but that should only result in slight changes in phase, which I guess would result in a higher BER.
Oh, and you forogot about cosmic rays 
-Steve