dslreports logo
site
 
    All Forums Hot Topics Gallery
spc

spacer




how-to block ads


Search Topic:
uniqs
798
share rss forum feed


espaeth
Digital Plumber
Premium,MVM
join:2001-04-21
Minneapolis, MN
kudos:2
Reviews:
·Vitelity VOIP

1 edit

1 recommendation

This isn't that difficult, people...

Drawing conclusions about access network capacity using numbers from consolidated transit links is a misguided approach. That's like saying there's no congestion on Las Vegas Blvd because I-15 is running free and clear.

The carriers have no problem handling the load because core transmission technology has continued to improve over the last decade. We've gone from OC3/12/48 circuits to GigE, then to 10GigE, now 40G OC768 and 100GigE interfaces are starting to come around. They can use the same physical cabling they started out with, and just swing the endpoints from GigE ports to 10GigE ports and get an instant 1000% boost in capacity.

There has been no similar evolution in either the DOCSIS or DSL delivery platforms. The 38mbps DOCSIS downstream channels are basically the same spec as when they started rolling out HSI over cable in the late 90s. There's been few significant changes to the DSL to improve speeds -- and distance is still the determining factor over your max data rate. DSL does have and advantage in that at least the connection to the DSLAMs can be upgraded to add capacity at the DSLAM level, but to increase capacity out to the subscriber edge means deeper deployments like what Qwest and ATT are doing. On the DOCSIS side, there's no "interface swap" solution -- the only thing that can be done at the moment is to keep doing node splits to get the subscriber per channel densities to continue to come down. DOCSIS3 has been claimed to be the savior of some of these problems, but even though the spec was finalized in 2006 there is still only a single vendor (Casa) with a fully certified CMTS today. Also, the largest benefits of D3 don't come about until you can replace the majority of your installed cable modems on the plant -- so we need to get new hardware in the homes of 10s of millions of people before the benefits of a 300-400% bump in backbone capacity will be realized. (all at significantly greater costs than the transit/carrier link upgrades where they get 1000+% bumps in capacity)

Over the last decade the strategy on the DOCSIS plant has been to set a limit, observe, split capacity where necessary, increase limit, return to "observe" step. What started out as 1.5mbps connections into a 38mbps channel has evolved to 16+mbps connections into a 38mbps channel.

The real problem we are facing is a shift in the type of traffic we are seeing. Most networks are designed around a "burst" utilization model; it's expected for user demands of the network to increase over time and for there to be periods of intense network activity. What hasn't been expected is the shift to long-duration heavy impact utilization brought about by 30+ minute streaming applications or P2P apps that generate traffic constantly.

To sell the situation as impending doom is a bit misguided. There are very few technology related issues that cannot be solved by a combination of time and money. In this case, if the trend for extended long-duration network impact is going to continue, then the provisioning model at the edge is going to need to change, and that's going to bring about new cost structures. At the carrier level things are set -- the carriers charge by either full interface capacity or measured utilization on an interface, so the more you use the more you pay -- the network is self sustaining to handle future growth. On the access side, however, if there will continue to be drivers to force drastic changes in the oversubscription ratios of the infrastructure then there will be a similar drastic change in the cost structure.



pizz
bye bye twc. hello Comcast.
Premium
join:2000-10-27
Astoria, NY
Reviews:
·Time Warner Cable

If there aren't any Vendors selling tech, its impossible to implement it. I 100% agree that replacing end-user modems mta/emta will be a PITA to do. But here is where you and I disagree.

I dont think the congestion issue or the bandwidth crunch/caps have anything to do with the network side of things. I think its the suits in the exec office's making this to be such an issue. Because they simply do not want to allocate funds to networking upgrades. They'd rather have their plant techs/engineers etc.. duct-tape and try and it together.

There are limitations on any network of what it can and cannot handle. Congestion does happen, but there is so many QOS tools available, that it should never be as bad as companies are making it sound.

If Plant Engineers/Techs etc.. were making the decisions, not the suits. I dont think the last mile will be an issue in 2008 and beyond. Nor do i think Caps or PIS, or whatever new scheme of the month is. The only way for MSO's to compete with Fiber or DSL is to spend the money on the network, so users dont leave it.



funchords
Hello
Premium,MVM
join:2001-03-11
Yarmouth Port, MA
kudos:6
reply to espaeth

Your message is spot on WRT to the last mile. The Exaflood prediction wasn't a last-mile prediction, though -- it was an Internet prediction. It focused on these long, middle-of-the-network links.



rawgerz
The hell was that?
Premium
join:2004-10-03
Grove City, PA
reply to pizz

Well obviously they don't want to spend money on upgrades, most ISPs are publicly owned. Investors want to see gains. Doing a massive upgrade with no real ROI is corporate suicide.

I think being a public company is part of the problem. You can't do a whole lot if you're constantly walking on egg shells.
--

You can't make all the people happy all of the time. But it should be common sense to shoot for the majority.



espaeth
Digital Plumber
Premium,MVM
join:2001-04-21
Minneapolis, MN
kudos:2
Reviews:
·Vitelity VOIP
reply to pizz

said by pizz:

I think its the suits in the exec office's making this to be such an issue. Because they simply do not want to allocate funds to networking upgrades.
There's definitely some of that going on, but that's a side effect of the pricing model. At the carrier level when they spend money to upgrade the network they get more capacity to sell to increase their revenue flow. On broadband networks with flat-rate pricing, they get the same revenue whether they invest $1 or $1,000,000,000 on network upgrades. Their only motivation is to maintain subscriber numbers, which fosters a network of mediocrity. They only have to be just good enough so most subscribers don't leave; there's no additional revenue potential to be gained from making the network really kick ass.

said by pizz:

There are limitations on any network of what it can and cannot handle. Congestion does happen, but there is so many QOS tools available, that it should never be as bad as companies are making it sound.
QoS is difficult on a shared network with no traffic tiers because there's no strict standard about who really has network "right of way." Comcast's new system for bandwidth management involves queuing the traffic of those who have been using the network the most below that of light users during times of congestion. Still, allowing the light users a chance at getting decent performance during times of congestion is still being met with disapproval. Users want a fluffy service where the performance is never bad for anyone.. ever. Unfortunately that goal is in violent opposition to the $50 price point most folks also want.