Cisco Passes the IPTV Test
Light Reading enlisted the European Advanced Networking Test Center AG (EANTC) to run an extensive end-to-end IPTV test on Cisco equipment. Short answer: Cisco passed, showing its gear can handle a 60,000-user POP and, by extrapolation, a 1 million-subscriber IPTV service. It also demonstrated some important features, including Call Admissions Control (CAC) and Video Quality Experience (VQE), and worked as advertised, according to EANTC.
The test supports Cisco's contention that its proprietary features, including CAC and VQE, can improve the IPTV experience. Indeed, such special features may be necessary to corral the complexity of IPTV. EANTC found such additions compelling -- exciting even.
The full test report is available here: Testing Cisco's IPTV Infrastructure. An LRTV video discussing the IPTV test is available here: Cisco Passes IPTV Test.
EANTC runs these kinds of installations for a living, helping out service providers with "proof-of-concept" tests. IPTV presents a particular challenge, though, because TV channels have to be delivered real-time while competing for bandwidth with video-on-demand and regular Internet service. (The full "IPTV" service tested included all those parts and a handful of high-speed business customers as well.) And while EANTC has tested IPTV before, the Light Reading test "was certainly a much bigger scale of proof-of-concept testing" than the lab is used to, says Carsten Rossenhövel, managing director of EANTC. "Normally, you try to save in terms of ports. Here, we had 120 DSLAMs and more than 200 TV channels."
First, EANTC drafted a request for proposals (RFP) to provide 1 million IPTV customers with voice, video, and Internet access. Under "video," the requirements included a package of 20 high-definition and 200 standard-definition TV channels, plus 22-Gbit/s worth of video on demand (VOD) service, representing 6,400 simultaneous users, per point of presence (POP). Cisco's first test was to answer the RFP.
Cisco came back with a fiber-to-the-node design using 16 POPs consisting of Cisco's 7600 series routers. The 7600s then connected to 120 DSLAMs per POP, with DSL lines connecting users' homes to the network. For the network core, Cisco proposed using two of its CRS-1 routers.
To run the actual tests, EANTC and Cisco built the core network and one POP of the 16. That was the main thrust of the test: to make sure the aggregation network could serve this size of deployment. (The DSLAMs and DSL lines were emulated on Spirent Communications plc equipment; no DSLAMs were harmed in the making of this report.)
EANTC has previously tested CRS-1s to the 1 million-customer level. (See Cisco's CRS-1 Passes Our Test and 40-Gig Router Test Results.) So Rossenhövel was confident that if one POP worked, the network could handle the full 16-POP case.
In the end, Cisco passed a series of tests related to QOS, network resilience, and video performance.
Making the grade
For instance, EANTC simulated various failures and fiber breaks, and Cisco's routers managed to restore TV service in 625 milliseconds, worst case. And in QOS tests, Cisco correctly prevented the TV streams from getting disrupted when other types of traffic threatened to clog the network.
Now, keep in mind that Cisco's was the only architecture under scrutiny. The results don't say that Cisco has the best (or only) IPTV package out there; rather, it just shows Cisco's network can survive in a deployment of reasonably large scale.
As one would expect in a carrier-grade IPTV deployment, some tweaks were necessary. EANTC initially settled on one version of software for all the 7600s but had to load a different version in order to run multicast failover tests -- a change that prompted router reboots. And the CRS-1 software chosen turned out to have an MPLS problem, which Cisco found, not during the tests, but in a situation with a real-life customer. Cisco had to write up a fix.
On the hardware side, two of Cisco's eight Gigabit Ethernet linecards turned out to be defective and needed replacing, but EANTC says this is too small a sample to be statistically significant for the Cisco fleet.
Bells and whistles
One of the more interesting parts of the test, according to EANTC, involved Cisco's special features. Cisco was allowed to pick two tests of its own. So, the company asked EANTC to check out two features: CAC and VQE. The results gave EANTC one of its strongest conclusions in the whole process: that vendor-specific improvements are essential because IPTV can't run "out of the box" using just IP and MPLS standards.
"In order to successfully implement and make money from these services, [carriers] must partner with a vendor that will do extra and sophisticated network design work, and that can supply additional technology to prevent network over-subscription and other problems," the report reads.
CAC is meant to prevent VOD requests from overwhelming the network. If one too many VOD requests come in, CAC will reject them in order to preserve the bandwidth needed for TV channels. It might tick off that one user, but the alternative is to risk degraded quality for everyone watching the broadcast TV channels.
VQE software resides in the set-top box and is meant to compensate for any packets lost on the DSL part of the network. It gives the set-top box the ability to request any missing video frames, preventing any glitches in the video stream.
There are always questions about how easily a test will translate to the real-world environment. But Rossenhövel deemed the proof-of-concept of VQE and CAC as quite important, pointing out that IPTV networks will likely continue to require vendor-specific features to achieve the carrier-grade stamp. He was left with the impression that CAC and VQE are quite useful and would perform well in a large-scale network.
The Bottom line? The test successfully demonstrated one of the largest simulated IPTV networks to date and confirmed the functionality of Cisco's proprietary features.
— Craig Matsumoto, West Coast Editor, Light Reading