A6826A - PCI-X Dual Channel 2Gb/s Fibre Channel Adapter Performance Paper for Integrity Servers

6
Chart 3b: 128 KB Sequential Write Throughput
0
200
400
600
800
1000
1200
1400
1600
Number of Ports
Throughput (MB/s)
rx2600
186 372 557 739 928 1111 1295 1480
rx4640
188 376 564 752 938 1124 1307 1486
rx7620
186 372 558 742 921 1103 1287 1468
rx8620
186 372 558 722 908 1094 1278 1461
Integrity-SD
188 375 560 751 938 1126 1313 1496
12345678
rx2600 with two 900MHz Itanium 2 processors and 8GB Physical memory
rx4640 with four 1.3GHz Itanium 2 processors and 8GB Physical memory
rx7620, 2 cell partition, each cell with four 1.3 GHz Itanium 2 processors and 8GB Physical memory
rx8620, 2 cell partition, each cell with four 1.3GHz Itanium 2 processors and 8GB Physical memory
Integrity Superdome, 2 cell partition with eight 1.5GHz Itanium 2 processors and 8GB physical memory
Chart 3b shows the write throughput performance of 1, 2, 3 and 4 A6826A adapters installed in rx2600,
rx4640, rx7620, rx8620 and HP Integrity Superdome servers. The X-axis represents the number of ports. The Y-
axis represents the throughput performance in MB/s.
Chart 3b shows excellent write throughput performance with linear scaling for 4 A6826A adapters installed in
rx2600, rx4640, rx7620, rx8620 and HP Integrity Superdome servers, reaching an aggregate throughput of
1496 MB/s.
The table below summarizes the maximum number of cards that can be installed in the tested servers while
maintaining linear scalability. Additional adapters may be installed to provide better connectivity.
Maximum number of A6826As for linear scalability
Ports operating
speed
rx2600 rx4640 rx7620 rx8620 Integrity
Superdome
Both Ports at 2Gb/s 4 4 2 per I/O
Cage
2 per I/O
Cage
2 per I/O Cage
One port at 1 Gb/s
and other at 2Gb/s
4 4 3 per I/O
Cage
3 per I/O
Cage
3 per I/O Cage
Both ports at 1 Gb/s 4 6 4 per I/O
Cage
4 per I/O
Cage
4 per I/O Cage