Printer Friendly

DSL's rapid growth demands network expansion.

Quality DSL-related services must be ensured.

The advent of DSL (digital subscriber line) service has fueled the demand for high speed Internet, intranet, and extranet access for small to midsized businesses and residential customers. SOHOs (small office, home office) generally prefer symmetric DSL (SDSL) since e-commerce, posting information to a local area network (LAN), or maintaining Web servers requires as much bandwidth upstream as downstream. Residential customers, on the other hand, are primarily interested in downloading information so, for them, asymmetrical DSL (ADSL) is sufficient. With businesses more willing to pay for customer service-level agreements and quality of service (QoS), the focus of DSL deployments is moving away from the consumer market to the. SOHO market.

As SOHOs command the greatest attention in the growing DSL market, they also demand new applications--virtual private networking, voice over IP (VolP), and video--that the DSL architecture must support. This growth in data traffic has resulted in larger demands on network equipment and on the increasingly complex networks themselves that are using multiple technologies (Ethernet to ATM, Frame Relay to ATM), adding newer mediums (like cable modems) and growing in size to hundreds--or even thousands--of ports. While these hybrid networks grow more complex, the very complexion of the network traffic is also changing. The converged network, transporting voice and video, in addition to data, has become the norm. In addition to handling ever-rising traffic, now the network must contend with the delay-sensitive nature of voice and video and incompatibilities between SDSL and ADSL. These are all very real obstacles to DSL deployment on a mass scale and to satisfying service-level agreements and ensuring QoS.

As the demand for DSL service escalates in the coming years, how will DSL deployments be able to keep pace with this demand? Although DSL has tremendous market value, its position is not yet secure. There is a constant threat by cable companies that aggressively deploy cable-modem solutions in an effort to switch telephone users from the traditional dial service offered by telcos. And new developments in broadband satellites and wireless will be competing for DSL's business as well.


What then will make DSL a success? DSL will triumph when the user experiences the benefits of high-speed access combined with always being "on" the Internet--as well as having remote access--particularly the ability to work while away from the office--to the LAN. To meet the user's need for continuous high-speed access, it is important to view DSL deployment from an overall network performance perspective. This developing complexity of the network has made it all the more important to test and forecast network performance. Network managers have come to realize that testing for cross talk, power consumption, line quality, and modem interoperability is only part of the quality equation. To make a DSL access network successful, the performance of the network needs to be verified as an end-to-end experience. Metrics need to be in place not only to accurately test a DSL network, but to validate that the user is receiving the expected level and QoS prior to actually turning the first user "on" to the network. As timing is key to a successful service launch, network service providers, as well as network equipment manufacturers, are in need of a simple, yet accurate, test methodology to rapidly qualify such a complex wide area networking (WAN) solution, minimizing surprises when networks go "live".

The one ubiquitous protocol to make this vision a reality is IP. Today, network service providers are scrambling to offer a high quality IP over DSL access network to rejuvenate their infrastructures. However, if the user expects high quality, like that of their current telephone system, then both equipment vendors, as well as service providers, need to recognize the underlying fact that end-to-end network performance is the key to success.

What, then, are the performance metrics that characterize a DSL network? What aspects need to be measured--data, voice, video, or a combination of all? Finally, how important is it to test DSL as part of the whole network?


Traditional testing of networks is focused on protocol analysis--troubleshooting what's broken--on a limited number of ports or on monitoring from a central site. These types of testing typically assess, for example, if the network is down or if any alarms exist. Both techniques are reactive and try to address the current status of the network and resolve problems.

True network-performance analysis addresses the "actual user experience issues," including how well a network works or "what if" scenarios, while stressing the network or device under test to its maximum. This type of testing is proactive and is performed on tens, hundreds, or even thousands of ports simultaneously. True network-performance analysis helps the equipment vendor qualify the solution for real-life situations, compare performance to expected design goals, and substantially reduce time to market.

For network service providers, true network-performance analysis allows comparison between competing equipment from various vendors and ensures the projected QoS before actually connecting customers to the network, minimizing the risk of revenue loss.


To judge real network performance, either the whole network, or a good representative part of the network needs to be put under test. For DSL environments, this implies stressing the central office multiplexer--the DSLAM (digital subscriber line access multiplexer)--with hundreds or thousands of streams flowing downstream towards the users while hundreds of other streams traverse the DSLAM in the opposite direction, from the users to the backbone network.

Some of the critical tests the network must undergo include; per-port frame/cell/ packet measurements, such as throughput, latency, packet loss, the capability of the equipment to maintain a committed information rate (CIR) and the ability to withstand traffic bursts (CBR).

However, finding the system that passes all these tests does not yet ensure true QoS. Modern metrics to determine the QoS go one layer up the stack to address individual streams and flows. Since QoS is a subjective matter, depending on the performance of the actual user's traffic flow, these metrics zero down on the user's IP traffic providing the means to repeat optimization tests when network parameters change.

In addition to these metrics, it is important to address the class of service offered to the user, such as measuring and tracking per-flow latency variation and tracking data for sequence consistency, as both these parameters may adversely affect applications, such as video and voice, Thus, a measurement of how well devices and networks optimize, prioritize, and segment these flows or streams of traffic is obtained.

Other factors of special importance to DSL access network topologies is the verification of IP multicast traffic behavior; characterizing the construction, deconstruction, and efficiency of IP multicast groups; firewall-device performance under load; and VPN load handling and efficiency in extreme loading conditions.

The next step in sophisticated performance metrics is addressing traffic shaping policies, such as DiffServ, RSVP, MPLS, or IEEE 802.1 VLAN (Virtual LANs) priorities.

Many network managers believe the magical tool to evaluate the network performance is extensive use of IP pings. Unfortunately this is a major mistake. The ping test provides round-trip latency only, while giving no indication of the delay at the remote end. Additionally, it provides no measurement of streams of data (i.e., no voice or video information) and no view of the network components. And, it definitely does not load the devices under test. The only true metric to evaluate the network latency entails one-way latency measurements, latency-variation measurements with no end-station delay, while the device is under stress conditions.

A focused test methodology should address the specific DSL performance testing needs, such as loading a DSL network on an end-to-end basis, allowing hundreds and even thousands of DSL modem ports to be tested in conjunction with single or multiple DSLAMs. Various device and network configurations need to be tested, such as Ethernet, ATM, or USB access ports and ATM, Ethernet, or Frame Relay trunk ports. Testing should support traditional and modern performance metrics to completely assess a full network solution, prior to deployment and actual connection to live users. A true network performance test solution should encompass the following metrics.

Throughput -- Determines the maximum capacity per connection (ATM VPI/VCI) with no loss, at various frame sizes and frame rates for the network or network device under test;

Packet loss or Goodput - Indicates the performance of the device or network under test in a heavily loaded state, by measuring the percentage of packets that are not forwarded due to lack of resources;

Stability over time - Measures the frame loss or percentage of line capacity for each supported connection (ATM VPI/VCI) over a prolonged period of time. Each sample is time-stamped in order to help determine the correlation with external events and to detect failed streams;

Cell latency and cell loss -- Determines ATM performance at the cell level;

Buffering test - Measures the buffer capacity of the device or network under test by sending bursts of traffic at the maximum rate and measuring the longest burst size (in total number of packets) at which no packets are dropped;

ATM integrity and ATM cell mis-insertion - Determines the accuracy of ATM cell transfer through the DSL-ATM network by evaluating the number of incorrectly received frames; and

End-to-end IP performance - Measures the speed and capacity of the device or network to forward IP traffic between two or more end points, utilizing the following metrics:

* Sequence Tracking - Checks if any frames were dropped (not received) or received out of sequence for each traffic stream;

* Latency per Stream -- Records latency for each traffic stream at a receive port;

* Latency Distribution -- Tracks latency distribution across various time intervals; and

* Latency over Time -- Records the number of frames received, minimum latency, and maximum latency, for each measurement interval.

Ultimately, the user experience is the only accurate metric in evaluating network performance. With the drive for high-speed Internet access, the stakes are high in the DSL-access game. Investments are huge, deployment and market penetration are full of potholes, with the greatest concern being, perhaps, the fierce competition by cable-modem service providers. The risks associated with not using proper metrics to evaluate network performance include: the potential network meltdown when the network is under stress--many users logging on at once; the unnecessary expenditure on inappropriate or unjust network products; the inability to predict or provide QoS required for customer applications; and the incomplete testing of new technology (e.g., using ping tests) which can lead to a false sense of security.

With the growth of competing technologies, one cannot always assume that the better technology will prevail. DSL is certainly the most promising technology for delivering high-speed Internet and remote access. However, performance-analysis testing is a must for ensuring DSL's deployment, quality of service, and overall success.

Circle 258 for more information from Netcom Systems, Inc.

Karoly is product marketing manager at Netcom Systems, Inc., Calabasas, Calif.
COPYRIGHT 2000 Nelson Publishing
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2000 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Title Annotation:Technology Information
Comment:DSL, particularly for the SOHO user, is facing challenges and competition.
Author:Karoly, Eran
Publication:Communications News
Geographic Code:1USA
Date:Mar 1, 2000
Previous Article:HDSL2: Ready To run.
Next Article:The evolution of DSL.

Related Articles
Eicon Technology Corporation Names New Vice President of Marketing & Business Development.
Mother, may I connect at higher speeds more cheaply?
The evolution of DSL.
Copper Mountain G.lite Solution Selected by Northpoint Communications as Part of Nationwide Rollout of DSL to Consumer Market.
Network Telephone is First to Offer Live VoDSL Using Lucent's PathStar--TM-- Access Server.

Terms of use | Privacy policy | Copyright © 2018 Farlex, Inc. | Feedback | For webmasters