Subscribe / Unsubscribe Enewsletters | Login | Register

Pencil Banner

How sourcing practices can reduce network latency

Bjarne Munch | Sept. 19, 2012
Enterprises that use Internet services in Asia for branch office connectivity continue to report latency issues. We outline what enterprises must do to minimize Internet latency and optimize application performance.

Enterprises that use Internet services in Asia for branch office connectivity continue to report latency issues. We outline what enterprises must do to minimize Internet latency and optimize application performance.

Common mistakes

Selecting an Internet service provider based purely on the cost of local Internet access service is a common practice, but this approach is causing application performance problems for enterprises that increasingly use the Internet to connect Asian branch offices to their data centers, and to cloud services.

Because enterprises focus on buying Internet access connectivity at their locations, and ignore evaluating whether Internet service can support their end-to-end connectivity requirements, they fail to ensure that the end-to-end latency that delay-sensitive applications require can be fulfilled. Thus, end-to-end network latency often is two to three times greater than what can be optimally obtained, and can render applications unusable.

Extended traffic

Poor application performance often is blamed on the lack of quality of service (QoS), and on an overly congested Internet infrastructure. However, in Asia, the most severe application performance issues often are caused by traffic taking an unnecessarily long physical path through the Internet.

This results in increased transport latency, which leads to unacceptably slow application response times, and bad user experiences.

One key reason providers route Internet traffic inefficiently is that they own limited infrastructure, and rely heavily on interconnectivity through other providers to extend their reach. Many providers use the lowest-cost routes and interconnections, which often leads to high latency. Thus, enterprises should not source Internet services based on cost alone, but on cost and performance.

Matching ISP selection with changing usage

Because enterprises no longer use Internet services merely to gain Internet access, but also as their company intranet, enterprise network managers must incorporate their specific connectivity requirements into the Internet service provider selection process to include end-to-end connectivity across the Internet for all office locations.

This will encourage enterprises to standardize on a limited set of providers that can reliably supply suitable bandwidth, network latency and packet loss among all enterprise locations.

Enterprises can chose overlay providers, such as Akamai or Internap, that use route optimization techniques to evaluate various route choices among specific endpoints on the Internet, and thereby select the lowest-latency options. However, this means that enterprises must source local Internet access services and Internet routing overlay services, which adds costs.

Several regional network service providers, including NTT Communications, Pacnet, PCCW, SingTel and Verizon, are beginning to offer regional Internet connectivity services. Because these providers control the end-to-end connections, several providers, such as NTT and Verizon, offer SLAs with performance guarantees for latency, packet loss and jitter, as outlined in their published standard offerings.

 

1  2  Next Page 

Sign up for Computerworld eNewsletters.