Understanding Network/Application Performance

If you’re experiencing application performance problems, check your network; it may be robbing you blind

You’re in a NASCAR race — driving one of the world’s fastest high-performance cars, zipping along at lightning speeds. Now take that race-winning speedster off track and drive it down a back-country road littered with rocks, pot holes, and a washboard surface. Same car… but certainly not the same high-performance experience.

Operating mission-critical applications over a faulty network often yields the same frustrating results. That high-performance application can’t run optimally over a network with performance-hindering potholes like duplex mismatches, media errors, suppressed bandwidth, black hole hops, router loops, MTU bottlenecks, and collision domain violations.

Application performance is absolutely defined by the quality of the network. A troublesome or failing network produces immediate and measurable consequences. For any enterprise, this is a productivity hemorrhage; for vendors, it’s a nightmare.

Unfortunately, very few companies know that their networks are stealing performance. And if you’re a network-dependent vendor, selling hardware or software that uses the network, not only do you have your network to contend with, you also have to deal with your customers’ networks.

Network managers spend 80% of their time finding network problems; only 20% trying to fix them

A computing environment consisting of thousands of network devices, computer hardware, and application software is typically no faster than its slowest component. However, finding those slowest components is difficult and time consuming.

The list of potential bottlenecks is almost endless. An inexpensive or mis-configured network adapter in an important server prevents that server from handling its share of the workload. A poorly engineered or mis-configured router, switch or hub keeps a network segment from running at full speed. A client/server application works well on a local network, but high latency across a wide area network will slow the application.

Using technology that specializes in locating network faults results in huge gains in staff efficiency and productivity. Without these tools, tracking down the causes of network bottlenecks is largely a manual, trial-and-error process. Simplifying the process allows expensive skilled staff to transfer their priorities from fault-location tasks to actually fixing the problems.

Furthermore, identifying and repairing network bottlenecks and congestion points allows organizations to improve performance troubleshooting efficiencies, and maximize the performance and return on investment (ROI) of IP networking infrastructures.

80% of network problems are identified by users complaining about application performance

Even with network monitoring tools in place, most performance problems aren’t detected by IT professionals until the phone starts ringing and the complaints pour in.
This situation contributes to worldwide vendor and enterprise support costs of more than $400 billion per year, or $2,500 per PC user per year (which excludes user productivity and business impacts).
Obviously, the benefits of improving this situation are enormous, including:
• Decreased cost of support staff and travel
• Improved customer satisfaction and end user productivity
• Improved employee retention
• Enhanced revenue and support margins

More than 70% of the time, system bottlenecks are related to a configuration or environmental issue rather than an application defect

When your customer is working with the application you sold them and it grinds to a halt, what do you think their first response will be?
a) “That darn network of mine is slowing things down again, I better get someone to look at it,”
or
b) “This expensive program I just bought is crap and I’m going to call them up right now and give them a piece of my mind.”

If you picked B, you’re probably right. While it’s most likely the network that’s impacting your application’s performance, you’re going to have a tough time convincing your customer of that. With multiple third parties, limited access to your customers’ networks, and an extremely complex infrastructure, diagnosing a network-based problem is usually time-consuming and costly. In Mercury Interactive’s experience, more than 70% of the time, system bottlenecks are related to a configuration or environmental issue rather than an application defect. Ignoring the problem is worse.

More than 50% of trouble tickets are generated by Fast Ethernet duplex-mismatch conditions, according to a study by the NASA Jet Propulsion Laboratory.

A duplex conflict or mismatch is the mixing of transmission rules between two network interfaces that are connected with each interface implementing a different rule set. This results in simultaneous attempts at transmission causing lost packets.

When multiple applications attempt to transmit simultaneously on a link with a duplex mismatch, packets are lost and re-transmissions are forced; applications begin to slow down and sometimes fail, particularly under high traffic conditions. This is a common scenario for application or FTP servers, and performance degradations of 50% to 80% from expected levels are quite common.

The bad news is in most scenarios a user is not able to check this problem without access to advance network technology or an experienced network engineer with a lot of patience on hand to find and fix it.
If you have a managed Ethernet switch, determine the duplex setting at the switch. The method in which you do this is dependant on your switch vendor. Then determine that the duplex setting of you local machine matches. If you have different vendor equipment at the switch and PC end, using auto setting is strongly discouraged.

If the problem lies within the network, a network element to network element connection, you need the service of your WAN network administrator, which is probably outsourced to a carrier.

When a duplex problem is suspected, you have to look in two places to properly determine the problem. You can imagine how this becomes an impossible task to be tackled proactively when you have thousands of machines. Further, Windows update may change the duplex setting on your PC, causing problems to crop up unexpectedly.

As you can see, the vast majority of network problems, such as duplex conflicts, are difficult to find but easy to resolve. This bodes well for support professionals if they are equipped with the right technology to deal with network problems. Or, very badly, if they’re not.

Think back to the NASCAR scenario. If you were the driver heading down the country road, would you pull over and start tearing apart your motor or would you look for a better road? The same common sense should be applied when application performance problems hit – start with the network.

This is the first in a series of articles submitted by Apparent Networks, Inc. to educate you about the importance of having a sound network. Next month, discover The Seven Deadly Sins of your Network: “What you don’t know about your customers’ networks can hurt you.” For more information about troubleshooting your customer’s networks, go to: www.apparentnetworks.com.