What is the performance overhead of the Balance 305 Rev2. I realize this could be interpreted in many ways. To be a bit more specific, consider a web server with a public IP address. How much latency would be added by putting it behind the 305? I am sure it would be less than 1ms. Would it be less than .3ms? About where would it fall. Also consider that the peplink cpu load is less than 50%.


It should be very very low, I cannot give a definite amount as this would vary from web server to web server and taking into account that the server is plugged directly into the Balance, I would think you would be on par saying less or close to 1ms.

The latency added would not so much be a factor of the web server’s performance as the added latency would be solely due to the peplink processing overhead. It would nice if there was some sort of published hard metric.


This will not happen if you choose a right model for your environment. Status of WAN link (quality, available bandwidth and etc…) is the bottleneck based on my experience.

The bottleneck is always the network–this is not what I am asking. I am asking about the latency being added by the Peplink.

Consider 2 test cases using the same hardware:

Case 1:
dc switch 1 –> web server with public ip address from data center
dc switch 1 –> desktop computer with public ip address from data center

Two computers connected a data center’s switch each with a public ip addresses. Now measure the latency between the computers (most likely less than 1ms).

Case 2: add the peplink:
dc switch 1 –> peplink 1 –> switch 2 –> web server with local ip address
dc switch 1 –> desktop computer with public ip address

The peplink and desktop computer are plugged into the same switch. The web server is sitting behind the peplink and gets a local ip address from the peplink. Measure the latency.

Case 1’s latency will be lower than Case 2’s latency as the peplink has a cpu and needs to make firewall and routing decisions (also switch 2 will be adding latency as well). The measurable difference might be milliseconds, microseconds or nanoseconds. Nonetheless, it would be higher.


In general, latency is caused by distance between the end to end device.

Like your test case 1 and 2. Distance in test case 2 is higher than test case 1. So latency in test case 2 will be higher. There is no regular way to tell latency caused by a gateway device.

Hope this help.

Latency is caused by distance traveled (1 mile is around 8.2 microseconds / 100 miles around .82ms) and switch/routher processing overhead.

Obviously if your equipment is using older CPUs and slower memory your latency will be higher than when using faster CPUs/Memory. Some switches are able to do hardware assisted switching with is much faster than software based switching. The latency produced by every router and switch can be measured–you must not be aware of these metrics.


We do not have these metrics for you. You may of course test within your own network. If you notice latency to be out of the ordinary, get a hold of our support team and they’ll take a closer look.