Your Flate-Rate Bandwidth Needs Monitoring – Here’s Why
The following excerpt is from Datacate’s Colocation Survival Guide – get the full Guide here.
Much like power usage, bandwidth usage is rarely a “hit the ground running” scenario; most colocations start out at a moderate usage level, and grow as demand grows over weeks or months. You know best which scenario is more applicable to your colocation, but in any case, you’ll need to keep an eye on bandwidth usage, and this applies whether you have a metered or unmetered bandwidth plan.
In the case of a metered plan, the need to monitor bandwidth usage is fairly obvious: you have the potential to exceed your metered allotment and incur additional charges. Monitoring trends in your usage will enable you to be proactive in addressing that possibility, whether your solution is to throttle usage (not recommended because of performance issues, more on that below), simply pay the overage, or increase your plan commit to get better pricing on the additional usage.
With an unmetered plan, there’s no immediate concern of a financial impact from increased usage. Instead, network performance is the reason for active monitoring. This is why: as bandwidth utilization approaches the maximum limit of the network interface, performance begins to suffer – latency increases, dropped packets and refused connections become more common. The utilization levels at which performance degrades will vary according to the underlying network hardware and topology, along with a host of other factors, so defining an exact number is not really possible. As a general rule, if utilization is routinely reaching levels of 90% or more, those accessing your content or services are probably seeing performance issues, at least some of the time. In that case, you’ll need to consider upgrading your bandwidth plan to increase the upper limit of your connection speed. One way to keep an eye on the effects of high utilization is via the various uptime monitoring tools that are available:
Rest assured that your provider(s) are monitoring the facility network and power uptime, and if an interruption occurs at the facility level, they are aware of it. However, if your server crashes, they will not know that unless they are providing you with individualized monitoring of your equipment or uplink. Likewise, if you trip a breaker on your power strip due to an overload, the facility may not know this (if you trip a breaker on their upstream PDU, they will know that).
Best practices on your part dictate that you arrange for active monitoring of your colocation for 24/7 availability. If your provider offers a monitoring service, you can take advantage of that, if it make sense for you. You can instead use one of the many third party network monitoring services, such as Pingdom or Uptime Robot. For redundancy you can use a combination of two or more services. Since power interruptions will necessarily take equipment offline, separately monitoring power availability is not strictly necessary.