Why Capacity Matters More Than You Think

by Alec Pinkham on September 05, 2017

There are plenty of varieties of network monitoring these days, and various ways it’s used. The reason for using it, however, is fairly singular: to see and improve end-user experience. Network monitoring can ensure that the network has been running well, is running well currently and will continue to run well in the future, regardless of corporate circumstance. Capacity monitoring is an essential piece of monitoring the entire end-user experience.

Network monitoring should provide IT teams with useful metrics data to help them understand end-user experience. The set of metrics that are best predictive of network issues are those related to the capacity of a particular connection. From our perspective at AppNeta, this is where you can really see what’s happening along a particular network path. Bandwidth gets tossed around a lot, but capacity can give you a lot more important information.

When the utilized capacity number approaches the available capacity of a link, it suggests congestion that could lead to data loss and increased latency. If you see the provisioned capacity change, then there is deviation in the maximum observed capacity on the link that should be investigated and brought to the ISP’s attention (either with an SLA in place or not).

For every application delivery path, AppNeta displays capacity measurements that indicate the amount of capacity that was used, the amount left available and the provisioned capacity for the link over the specified time range. The provisioned capacity is shown either as the maximum observed during that time range, or can be entered by the user according to any SLA set with their ISP. Let’s look more closely at the various types of capacity and why they’re important in your monitoring.

Available and Total Capacity

AppNeta calculates the available capacity of a link based on the average dispersion of a number of packet trains. The average is used because we are actively sending data over the link in the same way that an application would be, so our monitoring tool experiences the same average available capacity on the line. The total capacity is measured  in a similar manner with the minimum dispersion over a series of packet trains. To account for brief periods where packet loss would skew the calculation to 0 or 100% available capacity, AppNeta’s TruPath includes a variable to account for packets lost in the calculation for available capacity.

• Lower-than-expected capacity: Capacity is always an end-to-end measurement and reflects a network layer value, whereas bandwidth is a physical layer measurement. The theoretical maximum of the capacity measured will be below the number reported by your ISP.  That’s due to link-layer headers and framing overhead. Beyond these factors, your capacity will be based on the bottleneck of the network connection.

Utilized Capacity

Utilized capacity is a calculated metric based on the total and available capacity of your network connection.

• Oversubscription: If you notice high utilized capacity coupled with no increase in AppNeta Usage data, this can be a sign of oversubscription. You’ll want to follow up with your ISP about this kind of drop.

Provisioned Capacity

While other metrics in this list are based on observed values that change often, the provisioned capacity shouldn’t change much. However, that doesn’t mean that it’s always static. The provisioned capacity number will depend on your particular ISP and SLA status.

• With an SLA: If an agreement is in place between you and your ISP, then AppNeta lets you set the provisioned capacity to match the guaranteed end-to-end service level. Ideally this remains static, but if it dips below the set level for long periods, then there are a number of different issues that could be occurring.
• No guaranteed SLA: If no SLA exists, AppNeta recommends leaving the provisioned capacity option unset. You pay your ISP for a certain link speed, but the internet itself is best effort. What you buy from the ISP is generally not correlated to end-to-end performance observed on the internet. AppNeta helps you identify when the capacity is the root cause of a performance issue. While no SLA is in place, there is still an expected level of service from your ISP.

Take note of the many varieties of capacity. This metric is easily overlooked, but provides a lot of important performance monitoring and end-user experience information.

AppNeta monitors detailed capacity information from cloud provider networks, both ways.

This shows detailed capacity data from the AWS cloud provider network.