BACKGROUND

Introduction
This document is a follow-up to Technical Note 98-110 (Tuning BackHome!/ADSM in a TCP/IP over Ethernet environment).
It presents results of performance testing using K-10000 and S-70000 systems over Ethernet (10Base-T), Fast Ethernet (100Base-TX) and ATM.

The test environment
The test environment consisted of the following hardware and software:

• An Himalaya K-10000 system (with 4 processors) running NSK Guardian D47.00. A 3615 Ethernet controller was connected to CPU 0;
• A Nevada S-70000 system (with 4 processors) running NSK Guardian G06.00. A ServerNet/4E ServerNet/4E 4-port Ethernet controller was connected to the system. Only one port was used for testing.
• A Compaq DeskPro Pentium-II 266MHz machine (with 64MB of RAM) running Windows NT Version 4.0SP3 and IBM ADSM Server Version 3.0. A Compaq 10/100 Ethernet adapter was used;
• For the Ethernet testing, the machines were part of the same segment through a hub. There was no activity on the segment (except the test cases) while performing the tests;
• For the Fast Ethernet testing, the machines (DeskPro and S-70000) were connected through a crossover cable. Full duplex was used;
• For the ATM testing, the machines (DeskPro and S-70000) were connected through a FORE Systems ATM switch. For our ATM testing, we did not use the special socket library that would have allowed us to use 256K buffers (instead of 31K). This future change will probably increase the throughput.

In all environments, the native TCP/IP stack was used.
 

    Here is a diagram of this environment:

Methodology
The following methodology was used for testing:

• All the testing has been done standalone on the system with minimal activity (10BaseT) or no activity (100BaseT, ATM) on the underlying network segments.
• Each test was a backup of a full Guardian sub-volume containing 16 files, ranging in size from 1.5 megabytes to 4.5 megabytes, for a total backup set size of just over 50 megabytes.
• Before starting the actual testing, a sample test was run many times to ensure that the results were consistent. Once these sample runs confirmed consistency, all other tests in the suite were run once.
• The start time and end time of each backup request was taken from the BCOM log (from OPEN message to CLOSE message).
• The testing that took place over TCP/IP was done using the optimal parameters for both the TCP/IP stack and the Ethernet controller, as described in the Technical Note Tuning BackHome!/ADSM in a TCP/IP over Ethernet environment on Bulk Data Transfer (TN98-110). Also, QIOMODE ON was used for all TLAM processes.
• The CPU usage was averaged from the VIEWSYS monitoring application. Please note that the purpose of the test was not to do precise measurements of the CPU usage of each component involved in a transfer. This is why VIEWSYS (with visual input only) was used instead of MEASCOM.

The data is extracted from the disk through the standard Guardian BACKUP utility. Its output is fed to the BFSBKUP process that emulates a tape drive. BFSBKUP repackages the data and send its output to the ADSM Requestor, which is responsible of sending the data over to TCP/IP via a standard socket. The BCOM Monitor oversees the global operation whereas the Logger handles messages issued by all BackHome!/ADSM components.