Posted by: Ed Tittel
how ISPs turn performance issues around on their customers, what Time-Warner says causes Internet delays, when one LAN using another LAN for Internet access
My ISP is Time Warner Cable/Road Runner, to whom I pay about $50 a month for so-called “Turbo Internet” service (which advertises download speed of 15.0 Mbps and a Turbo with PowerBoost mode that can take those speeds up to 30 Mbps, and upload speeds of up to 3.0 Mbps). Most of the time, I’m very happy with the service that they provide to me. But sometimes, as is inevitable with any service that provides egress to a public and shared resource, things do bog down from time to time. When such slowdowns are intermittent and infrequent, I’ve been around long enough to know it’s my job to grin and bear it. When such slowdowns become frequent or chronic, I always contact Time-Warner to try to find out what’s up.
Lately, they’ve been employing a couple of tactics that drive me absolutely crazy. Let me describe them, and then I’ll explain why I’m convinced that these tactics are unfair, misguided, and designed to exploit the ignorance of most Internet users about the way networks work and behave.
Tactic 1: The Speed Test
Any time a Time-Warner/RoadRunner tech support person analyzes network behavior they rely on a special network speed test tool. For my part of the world that tool resides at http://speedtest.texas.rr.com/, and it usually produces results that look like this:
What’s wrong with this picture? Not much, apparently, except that the results don’t necessarily reflect the speed of access that Time-Warner/Road Runner provides to the Internet. This speed test accesses the internal Road Runner network, to which my home is directly attached (in fact, the HTML Title value for this Web page reads “TWC Austin RDC Bandwidth Speed Test” in my Web browsers). Thus, it represents a “best-case” value for the maximum upload and download speeds that any node on my home network can achieve going through the gateway and onto the Time-Warner network, irrespective of final destination.
Now let’s compare these values to those returned by the SpeedTest link that’s built into my Network Meter desktop gadget, located online at http://ip-address.cc/speedtest/. I don’t necessary endorse or recommend this tool, but it’s readily available to me because the link is integrated into the desktop gadget that I use to monitor my network connection (and performance and health values) on my Windows 7 PCs. Here are values recorded within seconds of completing the previous Road Runner sanctioned speed test (and I like them, because they do not reflect some of the pathology I’ve been trying to get Road Runner to explain to me recently):
Notice that the very difference that you see between the two reports is what I’m trying to get Road Runner to address — namely, why is my connection to the Internet (that vast realm of IP addresses outside the Time-Warner/Road Runner network domains) running slowly? In this case the difference is 25.48 Mbps inside, and 8.46 outside, or 13.02 Mbps (153% of the smaller value and 51% of the greater value). When I call Time-Warner, it’s usually between 12 and 25 Mbps inside, and under 3 Mbps outside, or slow enough to make my e-mail quit working reliably and to make many Web pages refuse to load.
Frankly, I don’t necessarily care how fast my connection to Time-Warner’s servers might be, since I don’t use them very often, except perhaps as a sanity check to make sure my premises equipment is working properly (if the connection to the Time Warner network were as slow as my connection to the Internet that would strongly suggest that my network was the source for such problems, rather than what’s happening on the ISP’s side of that interconnection). And in fact, that leads me directly to Tactic number 2.
Tactic Number 2: It’s your network/PC, stupid!
The last time I contacted Time-Warner tech support was last Friday, when my wife expressed her exasperation at not being able to access her Yahoo email or the Russian-language news sites she accesses every day (both of which are outside the Time-Warner umbrella, and thus subject to the “inside-outside” speed differences I describe in the preceding tactic). I spent a very frustrating half-hour chatting with a support associate who put forward the following assertions:
1. Because Time-Warner’s speed test levels were at or above guaranteed service level values, their network was behaving properly. I get the reasoning behind this, but my service is not called “Time-Warner Turbo local network access” it’s called Time-Warner Turbo Internet, and in fact the Internet is what I want to access.
2. The next questions from the support tech elicited the facts that (a) I had a LAN at my home and (b) that Time-Warner was neither providing nor supporting the boundary equipment (I have a D-Link DIR-655 switch/router/firewall/WAP device at the boundary, and a D-Link D2100-AP 802.11g WAP hanging off a NetGear 108GS switch in my office. FWIW, both appear to be behaving properly and the logs from neither device suggest problems with either one).
3. Because my machines were having problems accessing the Internet at higher (or in fact, acceptable) speed levels, the problems lay with my machines and/or my Internet gateway device. We explored the idea that the Web browser caches needed refreshing and that local PC firewalls were causing the problems, but I had to observe that with up to half-a-dozen machines running different versions of Windows 7 and different Web browsers, it was unlikely that all would suffer the same slowdowns because all their environments were different. It’s also not clear to me how a poisoned Web cache or misconfigured browser could deliver normal results for access to the Time-Warner network and slow results for Internet access without indicating the presence of some kind of bottleneck at the boundary between Time-Warner’s internal networks and the Internet, rather than between my network and theirs.
3. This led to the observation that Ethernet is a shared medium and thus, individual machines on my LAN would have only a fraction of total bandwidth available to them for Internet access. This explanation totally overlooks that only one or two nodes are likely to access the Internet at any given moment, and that the speed difference values between inside and outside exceed the ratio of machines making simultaneous access to the overall bandwidth available. And the same observation I made for the preceding item also applies here — namely, that if the slowdown/bottleneck straddled the boundary between my network and theirs, then the speed test results for Time-Warner local access and real honest-to-goodness Internet access would have to be closer, and service level guarantees not met for that local connection/access as well.
That was the point where I gave up, and resolved to write this blog to document what I’d experienced. I’m going to send it to a local manager at Time-Warner and ask for a comment. I’ll report back here on what happens, and find myself more than ordinarily curious to see what kind of response I’m going to get.
Frankly, I don’t see how my network can be at fault when it can access the ISP’s own network at or above guaranteed performance values. It’s only the next “big hop” onto the Internet where problems manifest, and as far as I can tell, that has to be their problem not mine. So when their support associate indicated that I should schedule a for-a-fee service call to have a technician come and troubleshoot my network for me, I kind of lost my cool and terminated the support chat session. Upon reflection, I’m reconsidering and my decide to bring in a technician to see what he or she can find — but only if Time-Warner will refund the charges if the network problems turn out to be on their side of the demark, not on my side, as I strongly believe will turn out to the case.