Hey all,
I've been thinking about the impact that iCloud (by Apple) will have on the Internet.
My guess is that 99% of consumer internet access is Asymmetrical (DSL, Cable, wireless, etc) and iCloud when launched will 'upload' obscene amounts of gigs of music, tv, backups, email, photos, documents/data and so on to their data centres.
Now, don't misunderstand me, I love the concept of iCloud, as I do DropBox, but from an Access Providers perspective, I'm thinking this might be a 'bad thing'.
From what I can see there are some key issues:
* Users with plans that count upload and download together.
* The speed of Asymmetric tail technology such as DSL
* The design of access provider backhaul (from DSLAM to core) metrics
* The design of some transit metrics
So basically the potential issue is that a large residential provider could have thousands of users connect to iCloud, their connections slowed because of uploading data, burning their included bandwidth caps, slowing down the backhaul segment of the network, and as residential providers are mostly download, some purchase transit from their upstreams in an symmetric fashion.
This post is really just to prompt discussion if people think there is anything to actually worry about, or there are other implications that I've not really thought of yet.
…Skeeve
My understanding was that the whole point of iCloud is to not upload but
rather use Apple's stored music files as long as you have them in your
library. You have a valid point however with other similar services, like
amazon's. But that's been out for a while.
--Andrey
From: "Skeeve Stevens" <Skeeve@eintellego.net>
I've been thinking about the impact that iCloud (by Apple) will have
on the Internet.
Aw, c'mon; what a boring Whacky Weekend thread... 
So basically the potential issue is that a large residential provider
could have thousands of users connect to iCloud, their connections
slowed because of uploading data, burning their included bandwidth
caps, slowing down the backhaul segment of the network, and as
residential providers are mostly download, some purchase transit from
their upstreams in an symmetric fashion.
IOW: The Tragedy Of The Commons. Yup; this is definitely going to be fun.
This is, I think, slightly different from the Netflix Instant arguments we
always have: the ones wherein it's pointed out that carriers aren't really
entitled to charge Netflix because they've made improper bets on their
capacity engineering for What Might Come Next:
Apple really had better have already *known* what the engineering of
consumer grade Internet looked like (or several high level product and
engineering people are dangerously underqualified for their jobs), so
as this problem gets worse, and I concur that it will, the result will be
that they bet wrong.
Gambling means that sometimes you lose. Alas, the costs won't be on
Apple.
This seems to be an ongoing situation: carriers discovering that they
also bet wrong on how to engineer the network: they've been making the
beams thinner and thinner, and then along came something reasonably
rational... that was heavy enough to break them.
Anyone betting carriers will stop gambling quite so hard, and build
networks the way John Roebling built bridges?
Cheers,
-- jr 'first woodpecker that came along...' a
That is only for musicŠ Photos will be the big killer, documents and
iDevice backups as well.
ŠSkeeve
My guess is that 99% of consumer internet access is Asymmetrical (DSL, Cable,
wireless, etc) and iCloud when launched will 'upload' obscene amounts of
gigs of music, tv, backups, email, photos, documents/data and so on to their
idata centres.
This is probably not goind to be any harder on your network that BitTorrent
and friends.
From what I can see there are some key issues:
You missed the *really* key issue.
The more people store data in the cloud, the more irate people are going to be
calling your help desk if you have an outage. We've already seen a few news
stories where a cloud service has whoopsied and lost data.
(And yes, I know that technically, the fact that Joe Sixpack made a poor choice
of backup/storage strategies doesn't impose added duties on you. But your help
desk is going to have a hard time explaining that to a pissed-off Joe)
Am I the only one who thinks iCloud style services plus a Cogent peering
dispute is a likely "perfect storm" scenario? 
Gambling means that sometimes you lose. Alas, the costs won't be on
Apple.
This seems to be an ongoing situation: carriers discovering that they
also bet wrong on how to engineer the network: they've been making the
beams thinner and thinner, and then along came something reasonably
rational... that was heavy enough to break them.
Anyone betting carriers will stop gambling quite so hard, and build
networks the way John Roebling built bridges?
Well put.
I find it hard to blame the users for using the network. That is what
they pay the provider for. Any implicit assumptions about _how_ users
should use the network are simply corners cut to make things cheaper.
Gambling.
I think the effect will be limited unless Apple give alot more space away for free. there arny many iphones/pads/pods with just 5GB
Neil
What would be obscene about that is from a design POV it would be a
waste of resources.
"Music" and "TV" content are from a small number of sources, and
there are a massive potential number of users.
What should happen is instead of transmitting large video files...
block checksums should be transmitted,
and only files that are completely foreign should be transferred.
Whereas everything else being "backed up" is just an assignment of
account access to existing blocks that would
already have been stored on the content servers.
And then also, a user storing 10GB of music would probably take
only a few megabytes of their account space,
once the "space used" is evenly divided by the number of users that
have that block saved,
since a majority of music files backed up would be file-identical
with material someone else had already backed up,
and identical to material already in the iTunes store (which they
could pre-seed their database with).
How would storage vendors otherwise sell de duplication. I mean you could make the application smarter but that wouldn't sell more spinning rust or licenses.
Regards,
Seth
If you're worried about the problem of tens of thousands of users
simultaneously trying to upload files to a "central point" then I'm
not the slightest bit concerned about the network as a whole. In this
circumstance, one of two things will happen and possibly both,
depending: either a) the users will screw themselves by flooding their
uplinks in which case they will know what they've done to themselves
and will largely accept the problems for the durration or b) (and far
more likely) the links apple is using will become flooded or the
systems overloaded in some way or another in which case the customers
will say, "MAN, this *SUCKS*" and likely whine at apple. Because the
nature of the traffic isn't much different than, say, a windows patch
release, the traffic won't be *all of a sudden* but will be spread out
over hours and days. The probability of it causing disruptions
anywhere but at the immediate source or within the near vicinity of
the desination is low, as I see it. IMO, the only ones who really need
be concerned are Apple's bandwidth prodivers because traffic will be
concentrating within their networks and especially in the nodes apple
connects to.
-Wayne
If you think that call traffic's going to *Apple*, either you're an optimist,
or I'm nutsabago.
Cheers,
-- jr 'shut *up*' a
In my opinion. Home networking (including personal clouds) have to change the brain damaged model of asymmetric tail technologies. Giving back the original peer-to-peer nature of networking the asymmetricity of the access technologies will not be tolerable in such a level (1:10) we have today. Maybe 1:2 should be more acceptable.
You don't have to worry bout this changes, but access provider cannot claim any longer 100MBps (while upload speed ~10 Mbps), but probably 60-70
Mbps (with upload ~ 30 Mbps).... They have to retune access services.
Best Regards,
Janos
The current apple "media" is reporting it's likely going to amazon and microsoft azure.
I've not bothered to look too deeply at dns and packet traces myself.
I'm guessing that all these things are going to hurt the DSL providers more than the docsis/fttx/pon based providers. Those folks have broader capabilities by pushing updated configs to the devices. The DSL based providers are more limited in my experience and likely to see a poorer ratio of up:down. SDSL was just not common enough, so most A/VDSL based providers have something like 15:1.5 whereas comcast (for example) has 22:5. I've seen the 22:5 service burst (or should that be buffer/manage the queue) to around 80Mb/s down in some cases.
This is something you are unlikely to see from a DSL provider unless the equipment is in-building.
- Jared
Two people making the same mistake: end-user support telephone calls don't
generally go to datacenters, do they?
Cheers,
-- jra
Mohacsi Janos wrote:
In my opinion. Home networking (including personal clouds) have to change the brain damaged model of asymmetric tail technologies. Giving back the original peer-to-peer nature of networking the asymmetricity of the access technologies will not be tolerable in such a level (1:10) we have today. Maybe 1:2 should be more acceptable.
I think a more fundamental question is why in 2011 we're stuck with statically
shaped asymmetric up and down. You can pretty dynamically shape *within* a given
direction to do just about anything you want to the traffic, but I don't know of
last mile access technologies that do that *across* the up and downstream. If it
were more like ethernet that doesn't have those artificial distinctions, this
conversation would be moot.
I recall the reason that DOCSIS is asymmetric is had a lot to do with how they
carved out spectrum of the analog channels -- and relegating upstream to slots
that weren't very good for those analog channels. That's been about 15 years
ago though and in the mean time the internet has sort of become important.
Now things may have changed -- I'd love to hear about it. Maybe Fred or
John Chapman can comment.
Mike
In my opinion. Home networking (including personal clouds) have to change the brain damaged model of asymmetric tail technologies. Giving back the original peer-to-peer nature of networking the asymmetricity of the access technologies will not be tolerable in such a level (1:10) we have today. Maybe 1:2 should be more acceptable.
I think a more fundamental question is why in 2011 we're stuck with statically
shaped asymmetric up and down. You can pretty dynamically shape *within* a given
direction to do just about anything you want to the traffic, but I don't know of
last mile access technologies that do that *across* the up and downstream. If it
were more like ethernet that doesn't have those artificial distinctions, this
conversation would be moot.
I recall the reason that DOCSIS is asymmetric is had a lot to do with how they
carved out spectrum of the analog channels -- and relegating upstream to slots
that weren't very good for those analog channels. That's been about 15 years
ago though and in the mean time the internet has sort of become important.
With dsl technologies like vdsl (flexible) or adsl (fixed 1/8 or 1/24) the total bandwidth (up+down) is not linear.
example:
adsl: 1mbit up, 24mbit down - total 25mbit
can not be used with 12.5mbit up/down.
at the co the noise is very high, as there are many lines in a bundle and the dslams "cry" with high signal levels into the lines.
Also the crosstalk is high.
downstream:
co-side: dslams send signal with high level + high level noise.
cpe-side: signal arrives damped, noise arrives damped -> signal to
noise (snr) is acceptable.
high bandwiths can be achieved.
upstream:
cpe-side: cpe send signal with high level, low level noise
co-side: high level noise produce crosstalk to damped signal
that arrives from cpe -> signal to noise (snr) is low
only low bandwiths can be achieved.
so dsl technologies, that use old, unshielded cables operate now at the maximum what the cable can do (up to 30MHz with vdsl2).
Higher speeds can only be achieved with better cables; like fiber or coax.
coax technologies use in oposite to dsl technologies no point to point links but bus technology to connect several customers to one head-end.
asymmetric bandwith -> more clients per head-end.
high-speed symmetric services can only be offered with new network types like fiber.
Kind regards,
Ingo Flaschberger
I'm not 100% certain and have no references to back it up but I recall
reading an article which described the Apple cloud music strategy as being
one where for existing identified music it merely stores a reference of some
kind against your account rather than actually storing an additional copy.
Presumably for the sake of sanity it would be implemented in the application
where it saves the end user the cost/time of uploading as well, if for no
other reason that said cost/time/cpu resource would also be a real cost to
Apple directly or indirectly.
Mumble-something about "even for your own music you have
ripped-not-purchased, pay $nominal-annual-fee and it magically becomes a
legal licensed object" (which obviously they did because then it becomes
something they have one single stored copy of with references on remote
accounts).
Phil P
The copper technologies of DOCSIS and xDSL are well established in
residential deployments and they are asymmetric by design. I don't think
near-symmetric speeds are on the CableLab's and Broadband Forum's short list
of future features. Even GPON is 1:4. As more fiber is deployed, I believe
deployments will eventually migrate to some variation of EPON where
symmetricity is built into the design. In the meantime it is what it is.
Somewhat tangential, has anyone graphed out the upstream/downstream ratios
of the various technologies (various generations of dial-up, DSL, fiber,
etc)?
Frank
Here's a very timely article on the topic of DOCSIS upstream:
http://accessintelligence.imirus.com/Mpowered/book/vcomm11/i8/p18
Frank