3Com SuperStack II PathBuilder S330 Reference Manual page 245

Wan access switches
Hide thumbs Also See for SuperStack II PathBuilder S330:
Table of Contents

Advertisement

The default memory configuration uses all available memory. In order to reallocate
memory to a given queue, you must first deallocate it from one of the other
queues.
Since the CTX is output-buffered, and if the WAN trunk has the lowest speed (for
nxT1/E1), the WAN trunk should take up most of the buffer.
CBR traffic (voice and video) requires the least latency during transmission, while
VBR traffic (email and FTP) is more tolerable to delays but more sensitive to errors.
The PathBuilder S330/S310 platform handles these traffic patterns with multiple
levels of priorities. Each switch port, except the first T1/E1 UNI port, has two
queues: one for high and one for low. The first T1/E1 UNI port or IMA group has
four queues: one for high (0), one for medium (1), and two for low (2,3) with
multiple traffic shapers.
To get a practical feel for the queue size configuration, consider an example in
which the PathBuilder S330 is configured on the trunk side with a single group of
4 x T1/E1. The other three groups on the trunk are not used. In this case, all the
cells coming from the high-speed OC3/STM-1 are going to end up in the T1/E1
buffers.
You could partition the memory on the trunk side as follows (with one block
equalling 256 cells and the minimum queue size being 3 blocks):
The T1/E1 buffer could be configured to 48K (192 blocks) cells large.
The high queue could be 768 cells or 3 blocks (normally small),
The medium queue could be up to 16 blocks large, depending on the number
of connections and the bursts tolerated.
The remaining 173 blocks could be divided evenly among the two low queues.
You could partition the remaining 16K cells (64 blocks) as follows:
3 blocks for expansion slot 7 high
3 blocks for expansion slot 7 low
3 blocks for OC3/STM-1 2 high
26 blocks for OC3/STM-1 low
3 blocks for SAR high
26 blocks for SAR low
This is just an example; normally the lower the speed, the more buffering is
needed. When the T1/E1 IMA is configured for more than one group, then the
first group gets four queues and the rest get two queues. In most cases, if the
groups have the same speed, then the cells reserved for the trunk can be divided
equally.
CTX Switch
237

Advertisement

Table of Contents
loading

This manual is also suitable for:

Superstack ii pathbuilder s310

Table of Contents