Unsolved

This post is more than 5 years old

1 Rookie

 • 

5 Posts

479

December 6th, 2011 01:00

DATA TRANSFER SPEED AT BACKEND

consider this situation: DAE 1:bus0 enclosure 0

                                  DAE 2:bus0 enclosure 1

A server  reading data from bus 0 enclosure 1 will take more time than reading from bus 0 enclosure 0. That is reading data from DAE2 to cache of the processor.

2 Intern

 • 

392 Posts

December 6th, 2011 04:00

And the question is?

Architecturally, all things being equal, on a quiet system with identical drives, and the same I/O access to Bus 0 Enclosure 0 (the DAE O/S) will be somewhat faster than any other enclosure.  However, it is not milliseconds faster, its in the order of microseconds.  This is because on a bus, the enclosures are 'daisy chained' together.  That is, 0_0 is directly connected to the backend port, 0_1 is connected to 0_0, and 0_2 would be connected to 0_1. Any transmission to 0_1 must traverse 0_0.  Short fibre channel cables connect the DAEs to each other.  The cables attach to Link Controller Cards (LCCs) within the DAE.  That several microsecond delay to 0_1 over 0_0 is due to the transmission speed through the additional cable lengths, and two LCCs.

EMC Unfied Storage System Fundamentals version 1.0 describes this for the CLARiiON in its 'Backend' section.  This document may still be available on Powerlink.  The latest version covers the VNX, which has a similar backend architecture, although implemented with the SAS protocol, and not Fibre Channel.  Fundamentals is available on Powerlink.

Top