Network Latency 101: How to Improve Latency in IT
As agencies' networks widen their reach through the adoption of cloud and mobile technologies, they face a major challenge in improving network latency. Sure, overall bandwidth is important, but high latency can destroy application usability, make streaming videos unwatchable and dramatically increase user frustration. For example, every drop of 20 milliseconds of network latency will result in a 7 to 15 percent decrease in page load times.
Here are some tips for improving network latency.
1. Realize that endpoints can impact network latency.
Users can run apps located anywhere across the Internet, on connections ranging from high-speed fiber lines to 3G mobile networks. To deal with this kind of variability, agencies should run monitoring tools, such as Ipswitch's WhatsUp Gold, to locate these apps and learn how they connect to the network. If necessary, an agency should boost its Internet connections to those locations.
2. Investigate the benefits of content delivery networks.
CDNs reduce many latency issues by having dedicated bandwidth and higher-quality links. They can help cut packet delivery times substantially in certain situations. But they aren't a solution to every problem; indeed, they work only for static web pages and other limited kinds of content. And they can be pricey, even for the smallest contracts or usage models. An alternative to CDNs is to boost quality of service agreements with existing Internet providers to improve latency.
3. Look carefully at network bottlenecks.
Network administrators should measure typical network traffic flows at different times of the day (right after lunch is a typically congested time). These measurements should account for the performance of the same protocols that the agency's applications use, such as HTTP and FTP. While many network engineers make use of measurement tools such as traceroute and ping, most modern applications and networks don't give much priority to ping packets, and admins need more advanced tools to understand HTTP behavior.
Among the steps agencies can take to reduce the latency of congested network nodes: upgrade the router or add more processing power or network adapters to a particular server. Network administrators can also reduce the number of nodes needed to traverse the enterprise. Finally, admins can consider centralizing network connection points to reduce transport latencies.
4. Ensure cloud providers are not contributing to network latency.
Agencies should understand how their cloud infrastructures are configured. And they must know where the particular pieces of network, applications, servers and storage fabrics are deployed and how they are connected to in-house equipment. Admins can take advantage of network auditing tools such as CloudSleuth and Stackdriver to help track down this information. CloudSleuth has attempted to mirror real-world app conditions and has set up a series of servers running simultaneous transactions from 32 locations around the world. The auditing service makes use of standard application servers from Apache Tomcat, Microsoft and Google with their sample e-commerce app. They then measure the response time of the application hosted at 17 cloud providers. Stackdriver can map the performance of particular pieces of a cloud software stack.
5. Make a direct connection.
Cloud providers have special services that allow agencies to directly connect their enterprise infrastructure with provider data centers. Agencies might consider using one of these options, such as Microsoft's Azure Virtual Network. While they can be pricey, these options can boost application performance and provide a more reliable latency across the Internet.
