Expand Configuration and Management Manual (H06.21+, J06.10+)

Tuning
Expand Configuration and Management Manual 529522-013
19 - 12
NAM Interface
NAM Interface
When a NAM interface is used, Layer 2 functions are managed by the NAM process,
thus reducing the load on the Expand line-handler process. Although the Expand line-
handler process has a potentially greater upper throughput limit when it uses a NAM
interface, overall system processor requirements are not reduced because some of the
workload is shifted to the NAM process. There is also an additional cost per packet for
the interprocess message between the Expand line-handler process and the NAM
process.
On a high-powered processor, such as an Integrity NonStop NS-series server, this
extra available processor power can allow an Expand line-handler process to drive
multiple high-speed lines and greatly extend throughput.
The relationship between the size of the Expand packet and the NAM network native
packet size has a major influence on the available Expand line-handler process
bandwidth. Each Expand packet passed to the NAM is handled as a message by the
NAM.
If the NAM network native packet size is smaller than the Expand packet size, the NAM
process must fragment the Expand packet. If the Expand packet size is smaller than
the NAM packet size, the number of messages to the NAM will be higher than if the
Expand packet size were the same size as the NAM packet size.
Data Compression
Data compression indirectly affects Expand line-handler process performance. By
shortening the length of a message, compression can reduce the number of packets
transmitted.
Because the data compression feature has an insignificant impact on the processor,
data compression should always be enabled unless you are certain that no data is
compressible. If compression is enabled and data is not compressible, data
compression actually causes messages to be slightly longer because the Expand
subsystem inserts a compression word every 255 words (510 bytes) of the message.
You can increase network efficiency by analyzing routine dat
a to determine the degree
of compressibility and then setting the frame size to carry the largest data-compressed
message. This technique can be a very effective way to economize processor
resources for point-to-point links with heavy, large-block message traffic.
Data compression is configured using the COMPRESS_ON modifier
.