EoPDH as an alternative for delivering rich media applications
By Serdar Kiykioglu, ANDA Networks -- 12/23/2008
Convergence of voice, video and data services are driving the growing demand for bandwidth, as mission critical transaction traffic and media-rich IP applications such as streaming video and peer-to-peer traffic are now transported over the same network connection. These new media-rich end-user applications demand resiliency, link error immunity, low latency and low cost, which imply the need for an infrastructure that offers bandwidth efficiency, scalability and reliability for the best user experience. Typically end users have employed expensive routers with WAN modules to accommodate Ethernet traffic from the LAN to WAN. This study highlights the use of Ethernet over PDH/SDH (Plesiochronous Digital Hierarchy/Synchronous Digital Hierarchy) based equipment that is based on GFP (Generic Framing Procedure) encapsulation and VCAT (Virtual Concatenation), LCAS (Link Capacity Adjustment Scheme) protocols, to cost effectively transport and deliver new, WAN based Ethernet connections as increasingly more rich media and latency dependent applications move over the WAN.
An evaluation of video services for enterprises and residential applications over a packet network was conducted utilizing ANDA EtherReach 2000 series, Ethernet over PDH (EoPDH) access equipment using GFP encapsulation and VCAT, LCAS protocols and Cisco 2800 access routers using MLPPP (Multi-Link Point-to-Point) based protocols, connected to identical video servers via standard T1 transport connections. Various latency dependent applications such as video streams, movies, and rich media content were tested over the network (See Table 1).
Two test setups were used to compare the performance of EoPDH and MLPPP for media service, which was delivered via Microsoft Media Server 2003. For the EoPDH testing, a streaming high definition movie was carried across a test network of four bonded T1s, consisting of ANDA's EtherReach 2000 series EoPDH equipment, in a book-ended configuration. The second setup carried the identical video and audio service simultaneously, across a test network of four bonded T1s via MLPPP via two Cisco 2800 routers in a book-ended configuration.
For the mobile operator, a major challenge is the cost of mobile backhaul, as it accounts for almost half the total operational expenses. To put this in a financial perspective, mobile carriers spent close to $19.5B in 2006 and are projected to grow their worldwide total mobile backhaul transport service charges to nearly $32.3B by 2010. (Infonetics, Wireless Backhaul Equipment Market report, 2007-2008.) It becomes critical for mobile operators to adopt solutions that are reliable, efficient and scalable for bandwidth use. This is important not only for the new, converged IP applications such as peer-to-peer streaming media, gaming and mobile TV, but also for the existing legacy media services over the mobile network. As the transport infrastructure is migrating to purely packet, layer-2 networks, running over services such as Carrier Ethernet, the links between the mobile base stations via the data transport network still require TDM connectivity for voice.
There are three methodologies used to bridge legacy TDM voice with newer IP traffic namely, using Carrier Ethernet over either wireline or wireless-based transport technologies, via: a) Bonding technologies with existing PDH/SDH based circuits; b) Pseudowire-based circuit emulation of voice-over-packet transport, and c) new synchronized-Ethernet and timing standards such as IEEE 1588 v2/v3. This article highlights results using current bonding technologies and their impact on delivery of multimedia traffic over the Ethernet transport network.
In this study, EoPDH and MLPPP based bonding methodologies for media delivery services have been tested to compare the performance from the user experience perspective, to show how EoPDH can deliver reliable, resilient services for applications such as streaming video and high definition TV that are particularly sensitive to link failure, latency and jitter.
When conducting the tests, engineers repeatedly introduced network failures such as interrupting T1 lines and causing network congestion by overloading the transmission pipe. It was verified that the media server application runs uninterrupted with only a degradation of video quality during some extreme test cases. However, with MLPPP, service interruption and service quality degradation occurred particularly in link degradation or failure situations where re-convergence of the connection must take place.
Test setup and methodologyBenefits of EoPDH
The engineers tested the media service delivery application across an in-building test network, using two setups via EoPDH and MLPPP. These test setups for the MLPPP and EoPDH are summarized in Figure 1 and Figure 2 respectively. The tests were designed to measure the user experience during streaming media transmission, under various network impairments, for the two setups.
The MLPPP service was delivered via Cisco 2800 routers with identical video servers and video display equipment, where the EoPDH setup was built around the ANDA ER2118 Ethernet over T1 network termination equipment. Media was delivered by a Microsoft Media Server 2003 and displayed by identical personal computers capable of displaying high definition video and high fidelity audio.
Prior to the subjective media streaming tests, engineers have tested both setups using an IXIA Network Tester, to establish a baseline on the performance of the setups via quantifiable metrics, such as throughput, latency and packets drop, for frame sizes 64, 128, 512, 1024 and 1518 bytes and for several iterations by stepping up the rate of ingress traffic.
For the subjective, user experience testing, engineers tested the EoPDH- and MLPPP-based services simultaneously, while the displays were placed side by side for comparison of the media quality while causing network impairment conditions on both setups, simultaneously, including interruption of T1 links such as:
- Interrupting and restoring 1 x T1 WAN link
- Interrupting and restoring 2 x T1 WAN link
- Interrupting and restoring 3 x T1 WAN link
- Interrupting and restoring 4 x T1 WAN link
All test subjects have concluded that the EoPDH-based setup yields a trouble-free viewing experience, since unlike MLPPP, EoPDH does not need the re-convergence times to recover service upon link failures or line degradation. With the EoPDH setup, media streaming was not interrupted when the WAN based bonded T1connections were interrupted, up until the last T1. Conversely, using MLPPP, each T1 interruption caused the stream to stop until the re-convergence of the link had been established.
The stuttering video and audio service and the temporary loss of service during complete or partial loss of WAN T1 link paths have been reported as annoying by the subjects, while the uninterrupted, graceful service recovery of the EoPDH based streams had been ranked better in terms of perceived quality and user experience.
Although the parametric measurements may indicate an advantage for MLPPP for the particular setup here, from a purely user experience perspective, EoPDH was ranked more suitable for the emerging IPTV and voice delivery services, thus revealing the cost-effective nature of EoPDH and the Ethernet demarcation products versus the advanced and more expensive routers.
Resilience
Based on IETF RFC 1990, MLPPP bundles serial links that run the PPP protocol for purposes such as higher bandwidth, link redundancy, IP address conservation, layer-2 load sharing and sequential transport of packets that have been fragmented. However, it lacks the link restoration capability that is expected from a carrier-grade network. For MLPPP to restore transmission, higher-layer protocol intervention is required which results in lower network availability and visible convergence gaps noted by end users in the video stream.
EoPDH on the other hand, is based on a pure layer-1 and 2 based protocols using the same concept of Ethernet over SONET/SDH and built upon proven technologies:
- EoPDH uses GFP (Generic Framing Protocol) as the method of encapsulation;
- EoPDH uses VCAT (Virtual Concatenation) to concatenate the required PDH connection for the required bandwidth of the transmission;
- EoPDH uses LCAS (Link Capacity Adjustment Scheme) for the adjustment of link capacity dynamically.
These standards enable EoPDH to not only respond to error conditions on the WAN links quickly, but also automatically adjust the pipe size in a hit-less and graceful re-convergence similar to what is done over the SONET/SDH networks today. MLPPP's link capacity adjustment methodology, while exhibiting similar or better test scores on test-set line measures (i.e.. throughput, jitter and delay) causes noticeable service disruptions (complete service interruption or stuttering) of the video streams. Conversely, EoPDH gracefully maintains the video signals (lower resolution video streams but no complete loss or interruption) during link degradation or link failures, thus maximizing end users' perception of service quality, particularly when viewing streaming media such as movies, sports, live broadcast media or participating in interactive gaming applications.
Why the differences? Independence from higher layer protocols
In the event of link failures, VCAT/LCAS provides the required hitless restoration capability for the transport without the need for any higher-level protocols, thereby minimizing convergence times and eliminating complete service disruptions. Using MLPPP, a failure of a fraction of the TDM links will cause the service to be interrupted until the higher layer protocol intervenes and restores the service, which may cause complete screen refreshes and/or major stuttering during the re-convergence of the transport links.
GFP supports CRC for error detection and correction, while PPP can detect but cannot correct errors. GFP also supports frame delineation and synchronization mechanisms. These capabilities help reduce the need for re-transmission due to error conditions such as bit errors and enable higher bandwidth efficiency. This advantage becomes noticeable in TCP traffic, since TCP treats bit errors as congestion.
It is the hit-less restoration capability of EoPDH that virtually eliminates the need to reserve bandwidth or establish a protection link, which is a significant savings in network equipment and infrastructure costs compared to an MLPPP based service delivery for a similar user experience.
EoPDH offers predictable bandwidth; it uses GFP with fixed overhead, while MLPPP's overhead is variable, which is proportional to the number of fragments per packet. This makes EoPDH more efficient for bandwidth utilization.
GFP also supports the ability to send management frames, which can be used for OAM, while for MLPPP, Link Control Protocol must be introduced in order to transport management frames. This requirement further lowers the efficiency of MLPPP.
EoPDH has the ability to transmit latency-sensitive traffic successfully, while in MLPPP buffering and realignment are needed since IP packets are fragmented into smaller packets and distributed over multiple links. Therefore, EoPDH offers lower latency.
With EoPDH's VCAT and GFP combination it can provide byte-level granularity of data, which reduces latency and jitter by reducing the buffering needs.
EoPDH endpoints typically do not require expensive and extensive buffering and memory within the equipment as they operate as pure layer-1 and layer-2 transport termination devices to maintain the connections and perform re-convergence using traditional SONET/SDH protocols. MLPPP however, operates at higher layers and is typically implemented in more expensive routers and WAN switching cards, where memory, buffering and specific higher layer QoS schemes can be accommodated to handle the re-convergence times necessary for multimedia and other latency dependent applications such as IP video and VoIP. For example typical EoPDH demarcation equipment can start from as little as $500 whereas typical WAN modules on routers alone typically start at $750 with the MLPPP based router we tested configured at roughly just under $5000.
Among competing transport bonding technologies, MLPPP and EoPDH both can multiplex bandwidth to deliver higher speed Carrier Ethernet based WAN services, however, EoPDH has advantages based on the types of services being supported, particularly where hitless bandwidth adjustment, low latency and efficient use of bandwidth are critical in multimedia applications such as streaming media, interactive gaming, or broadcast TV.
Service differentiation is a key attribute for today's competitive carriers as they move to offer more than just "fat pipes" over their network infrastructure investments. The potential ARPU per customer and overall customer retention becomes critical imperatives for both wireline and wireless mobile carriers as they move from basic Carrier Ethernet 1.0 based transport services to Carrier Ethernet 2.0 managed services, where customers are increasingly transporting media-rich content and mission critical applications all over a carrier provided, managed Ethernet service.
| Author Information |
 Serdar Kiykioglu is a product line manager at ANDA Networks, responsible for outbound marketing activities. He received his master's and bachelor’s degrees in electrical engineering from Istanbul Technical University (Istanbul, Turkey). You can reach Kiykioglu at skiykioglu@andanetworks.com. |
© 2009, Reed Business Information, a division of Reed Elsevier Inc. All Rights Reserved.