Quantum 3.5.1 Manuel D’Utilisation

Distributed LAN Server/Client Network and Memory Tuning

StorNext User’s Guide

331

* transfer buffer count
* transfer buffer size)

For example, suppose a Windows Distributed LAN Server is serving 

four file systems to 64 clients each using two NICs for data traffic. 

Also assume the server uses the defaults of sixteen transfer buffers 

and 256K per buffer. (On Windows, you can view and adjust the 

transfer buffer settings using the Client Configuration tool’s 

Distributed LAN tab.) Given this configuration, here is the result:
Required memory = 1GB + (4 * 2 * 64 * 16 * 256K) = 3GB

If not all clients mount all of the file systems, the memory 

requirement is reduced accordingly. For example, suppose in the 

previous example that half of the 64 LAN clients mount three of the 

four file systems, and the other half of the LAN clients mount the 

remaining file system. Given this configuration, here is the result:
Required memory = 1GB + (3 * 2 * 32 * 16 * 256K) + (1 * 2 * 32 * 16 * 

256K) = 1GB + 768MB + 256MB = 2GB
The calculation also changes when the number of NICs used for data 

traffic varies across clients. For example, in the previous example if 

the clients that mount only one file system each use three NICs for 

data instead of two, here is the result:
Required memory = 1GB + (3 * 2 * 32 * 16 * 256K) + (1 * 3 * 32 * 16 * 

•

 For a Linux Distributed LAN Server, use the following 

formula:
Required memory  = 1GB +

(# of file systems served
* # of NICs on the Distributed LAN Server used 

for

    Distributed LAN traffic
* server buffer count

This example assumes that a 64-bit version of Windows is 

being used on the Server. 32-bit Windows Distributed 

LAN Servers are restricted to small configurations using 

16 or fewer connections.