Wireless Versus Powerline: The Apple Router May Be The Rogue This Time
Last weekend, Northern California got some much-needed rain, so since I was stuck indoors, I decided to continue my experimentation in search of a root cause of my high definition video streaming woes. I couldn’t implement Mr. Write’s suggestion to ‘take Windows out of the equation’, since my system’s fundamentally based on a Media Center server and Extender infrastructure. But his comment got me wondering if what I was seeing glitch-wise was 80.211n inter-channel contention-related after all, or instead was somehow related to a Windows Vista TCP/IP flaw or random system resource starvation (CPU or otherwise) at my Dell laptop server.
I fired up the Network Performance Tuner utility and, whenever I saw a bandwidth glitch, glanced over at the CPU Meter Gadget sitting in Windows Vista’s Sidebar. I’d expected to see CPU usage spike up whenever bandwidth plummeted, if server-side system resource limitations were to blame. What I actually saw, however, was that the bandwidth dips coincided with timeframes of excessively low CPU utilization. Launching Task Manager confirmed my initial observations, and subsequently drilling down into Resource Monitor produced even more curious data; although the periodic network constraints had no correlation with HDD activity (another potential variable I was wondering about), they correlated with upward spikes in the CPU’s operating frequency.
So was inappropriate clock throttling the culprit? As it turns out, no; switching the laptop into a ‘high performance’ power management mode where the CPU always ran at peak clock speed didn’t fix things, and I still saw the relationship between bandwidth dips and CPU utilization dips. After thinking for a minute, I realized that while I was observing correlation, the CPU behavior wasn’t the cause of the network bandwidth constraint result…in fact, the converse was the case. Recall that the system’s networking stack is completely software-driven. Whenever the system doesn’t get ‘acks’ to the packets it’s sending out, CPU usage temporarily drops as the PC sits there waiting for a response. And when the response never comes (i.e. times out), indicative of dropped packets elsewhere in the chain, the CPU ramps up its clock rate as the Windows networking software stack launches into a re-attempt recovery mode.
My core question remained; what was causing the bandwidth dips in the network interconnecting the laptop and DMA2100 Windows Media Center Extender? I’d done a pretty thorough 5 GHz spectrum scan, but there was still the chance that a rogue SSID-broadcast-disabled access point, high-power cordless phone or C Band weather radar setup was causing spectrum corruption. My buddy’s 802.11n-supportive Apple Airport Express arrived in Saturday’s mail (on my birthday…how perfect is that?), so I decided to add it to the testing mix. Like my Apple router, it doesn’t support bonded channels in 2.4 GHz mode, but I figured that using would at least let me see if 5 GHz spectrum interference was the root cause of my wireless woes.
Over the weekend, I tried innumerable combinations of the:
- Dual-band-capable access point built into the Apple router, in both 2.4 GHz and 5 GHz operating modes
- Dual-band-capable Airport Express access point, in both 2.4 GHz and 5 GHz operating modes, and
- 5 GHz-only NETGEAR access point
When one or both external access points were in operation (i.e. in the latter case, when I bypassed the router’s built-in wireless capabilities), I also tried both connecting one or both of them to the router’s built-in switch, and one or both of them to the SMC external switch. And I always ensured that, in either the 2.4 or 5 GHz mode, the two access points’ wireless signals didn’t overlap and therefore didn’t destructively interfere with each other.
The bottom line and still-baffling results I’ve observed before remained stubbornly consistent, regardless of what equipment, frequency and connection variable set mentioned above I employed:
- Whenever I’m streaming using two access points interconnected by a wired switch (either built into the router or external to it), I see substantial, random glitches
- Whenever I’m streaming using only one wireless access point (either from the laptop to the router or from the router to the DMA2100), with either CAT5 or HomePlug AV handling the other half of the interconnect topology, I don’t see glitches
The Apple hardware doesn’t provide me access to status logs or screens that would, for example, report dropped-packet, collision, and other error statistics. Fortunately, the NETGEAR access point gives me this data, for both the wired and wireless ports:
Using it, I’ve been able to ascertain that the problem’s not with the wireless portion of the networking chain, which in fact seems quite robust; the dropped packets happen within the wired span between the access points. That conclusion points the blame finger squarely at my LAN nexus, the Apple router. Why does it seemingly struggle so much when managing the traffic between two wireless access points, including but not limited to its own, whereas it handles CAT5 and HomePlug AV peripherals with ease? If I could reliably stream high-definition video from source to destination within the LAN using only a single 802.11n bonded channel, I could avoid this router quirk, but my past testing has already shown that sustained bandwidth is insufficient in this particular scenario.
I’ve got a second-generation Apple Time Capsule sitting here, based on a third-generation Apple router design (versus my second-generation unit), which I’ll try next. I’ve also got two other 802.11n-capable routers:
Unlike Apple’s routers or HDD-inclusive Time Capsules, I can’t use the D-Link and Linksys gear in conjunction with a USB-tethered external hard drive to implement Time Machine backups (of questionable wireless robustness, but I digress). However, my Infrant/NETGEAR ReadyNAS just gained access to a firmware upgrade that’ll let it alternatively act as a Time Machine repository, so I’ve got other backup options even if I end up going with a non-Apple router.
I’d welcome reader comments on if you’ve seen similar issues in your dual-wireless access point networks, and how you’ve been able to surmount them.