Matching policies on inbound and outbound

[jbabio]

Is it a good rule of thumb to match inbound and outbound rate policies or another settings for that matter? I of course understand the difference of traffic direction but I notice a difference when I apply a priority policy of 4 on http inbound and outbound. If I only apply it to inbound it seems ok but when I include that same policy to oubound it is faster. This is standard web surfing pulling web pages from web servers out on the net. I was just curious how many do their settings of inbound vs outbound.

[Norbert]

Hi,

For simplicity I always try to use symmetric policies.

In your example you bump the Priority for outbound from Pri(3) to Pri(4) as the Pri(3) is applied by the Default Policy (Every flow through a PS gets a Policy when Shaping is on. Always!). If there is contention at Pri-level 3 and you lift the outbound request to level 4, you will see improvement.

Norbert

[jbabio]

Thank you for your insight. I was just curious as to how others handle their packeteer configs beings i never had any formal training. Thanks!

[paul.blitz]

My shaping philosophy?

1) work out the traffic that is causing problems. It is either big (>10% on the pie chart) or bursty (peak >70% of max b/w, as seen on the monitor page). Put it all into a folder, put a capping partition on the folder (10% guar, max 60 to 80%). For each class of traffic, apply a rate policy, probably rate 1, 2 or 3, depending on its relative importance / whether it is interactive. Same setup for inbound and outbound. This all gives packetshaper a good ability to control that traffic.

2) work out your important traffic. put it into an "important traffic" folder, with a very positive partition, then apply positive policies (rate / priority as suitable) per class.

3) Leave the "also ran" traffic alone.

Someone was asking about having different policies inbound to outbound. If I apply a policy, I do it for both IN and OUTbound, usually the same policy. In some cases, mainly for Citrix traffic, I'll use a priority on the client to server traffic, and a rate policy for server to client. Occasionally (if it's that important) I'll do the same for HTTP traffic, but not very often.