• United States

Are your bandwidth usage models wrong?

Mar 28, 20064 mins
System Management

As service providers plan their networks and make strategic technology choices – like fiber to the home vs. fiber to the node vs. enhanced CO DSL – they rely upon certain assumptions about how customers will want to use bandwidth, and what sorts of bandwidth requirements customers will have, and – most importantly – what minimum requirements users will seek from their provider (or go to another provider for).

The orthodoxy here is that certain applications like telco TV will use the majority of bandwidth provided to network customers, and that this bandwidth will be relatively predictable. The major U.S. telcos (Verizon and AT&T) are in the midst of massive network buildouts that are based on this assumption.

Likewise, there continues to be an assumption that customers will be consumers of bandwidth, and that an asymmetric skew towards downstream bandwidth is the right model. In the U.S., both FTTH and FTTN/VDSL deployments are being built around this assumption.

And at a high level view, these assumptions seem to be pretty correct. After all, TV is the biggest bandwidth consumer in the models that telcos use for these decisions. And TV does, at first glance, appear to be a predictable (and primarily downstream) application.

The monkey wrenches that most good models account for are peer-to-peer applications (like Bit Torrent), and other video services that bypass the telco TV offerings (like Google Video, Movielink and Apple’s iTunes Music Store). Most model builders and analysts understand the gigantic impact that these apps have on the network – particularly peer-to-peer today, but also the emerging impact of entertainment bypass services like those mentioned.

But we’ve seen some evidence recently – most notably in a Japanese research study conducted by IIJ Research Lab (Internet Initiative Japan) – that makes us think that most folks in the service provider community have underestimated the impact of these services. That’s because most folks believe that the massive impact of peer-to-peer on their networks is caused by a handful of “bandwidth hogs” who spend all of their time grabbing the latest high def torrents of Lost and 24, and then reseeding them back to other peer-to-peer users.

What the Japanese study, which was based on observation of over 4 million FTTH as well as DSL customers in the country, showed was that these high bandwidth consumers (called “heavy hitters” in the study) were not relegated to just a small subset of users. There was, in other words, no distinct user segmentation or demographic that defined the heavy hitters using P2P and other bandwidth intensive entertainment services on their FTTH connections.

Instead, these users tended to be distributed across all segments of the FTTH customer base. Some, according to the study, are users of P2P and other services who move to FTTH for the increased performance it gives them, but others are more casual users who upgrade to FTTH and then “look for applications to use the abundant bandwidth.”

In the Japanese study, the heavy hitters were still a small-ish segment of the overall user base (4% of users accounted for over 60% of the bandwidth consumed), but the usage trends of these users were not significantly different than those of more casual users. In fact the analysis found no clear boundaries between heavy hitters and normal customers – meaning anyone can become a heavy hitter, and probably will as peer-to-peer and entertainment bypass applications become more mainstream.

The takeaway we see here is the old Field of Dreams quote: build it and they will come. The only thing that really stood out about the heavy hitters was the fact that they were much more likely to be found amongst FTTH users – rather than DSL users. When users got access to higher bandwidth, they used it. When they got access to symmetric bandwidth, they used it symmetrically.

The bottom line is that it’s not just a handful of hardcore file sharers who can end up impacting your network and using tons of bandwidth. It can – and probably will – be any of your users. The use of bandwidth-intensive peer-to-peer and entertainment apps isn’t going to slow down any time soon, so you’d better have a plan to support and monetize these apps.