User manual MAXTOR 339MDWP1

[. . . ] 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 The Matrox NS-FNIC/4 Performance Numbers Under Red Hat Linux 6. 2 Matrox Networks November 6, 2000 Version 1. 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 Table of Contents Product Presentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 Benchmark Utilities Description. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Detailed Benchmarking Tests and Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 Test #1: Throughput and Server CPU Utilization. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . [. . . ] Sniffers are used to debug and monitor networking problems. TA B LE 1: SERVER AND CLIENT COMPUTER DESCRIPTIONS Server 1 Matrox NS-FNIC/4 700MHz Pentium III AGP Matrox G200 Client 1 Client 2 Client 3 #3) (Used in Test #3) · · · · Intel EtherPro 10/100 Intel EtherPro 10/100 3Com 3CR990-TX-97 166MHz Pentium 100MHz Pentium 700MHz Pentium III Matrox MGA PCI Matrox MGA PCI AGP Matrox G200 Millennium Millennium 128MB of RAM 64MB of RAM 64MB of RAM 128MB of RAM 20GB Hard Disk drive 1GB Hard Disk drive 1GB Hard Disk drive 20GB Hard Disk drive RedHat Linux 6. 2 RedHat Linux 6. 2 KDE RedHat Linux 6. 2 KDE RedHat Linux 6. 2 KDE (Kernel 2. 2. 16) workstations. workstations. 2 Benchmark Utilities Description Performance Numbers Under Red Hat Linux 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 DETAILED BENCHMARKING TESTS AND RESULTS TEST #1: THROUGHPUT AND SERVER CPU UTILIZATION The first test used the NetPerf benchmarking utility to obtain the throughput as well as the server CPU utilization. A packet size of 8 kilobytes (8192 bytes) was used throughout the test. The test bench was set up as shown in Figure 1, below. Figure 1: THROUGHPUT AND SERVER CPU UTILIZATION - BENCH CONFIGURATION Linux PC with a Matrox NS-FNIC/4 Linux PC with a 10/ 100 Ethernet Card Subnet 1 Subnet 2 Subnet 3 Subnet 4 Linux PC with a 10/ 100 Ethernet Card Linux PC with a 10/ 100 Ethernet Card Linux PC with a 10/ 100 Ethernet Card For this test, the clients were all 166MHz Pentiums. The results for throughput and server CPU utilization are summarized in Table 2. The throughput is in megabits per second (1 million bits per second) and the CPU utilization is in percentage. TA B L E 2: THROUGHPUT AND SERVER CPU UTILIZATION - NETPERF BENCHMARK RESULTS Port 1 Port 2 Port 3 Port 4 (Throughput / (Throughput / (Throughput / (Throughput / CPU Utilization) CPU Utilization) CPU Utilization) CPU Utilization) 10Mbits/s Half duplex 10Mbits/s Full duplex 100Mbits/s Half duplex 100Mbits/s Full duplex 6. 86 / 1. 85 9. 38 / 2. 59 65. 98 / 17. 5 88. 81 / 23. 97 6. 87 / 1. 86 9. 38 / 2. 51 66. 14 / 17. 06 88. 37 / 22. 98 6. 89 / 1. 85 9. 38 / 2. 59 66. 02 / 17. 51 89. 43 / 24. 33 6. 89 / 1. 85 9. 38 / 2. 66 66. 03 / 17. 56 89. 51 / 24. 11 On average, a Matrox NS-FNIC/4 port delivers a throughput of 89Mbits/s with 24% of server CPU utilization at 100Mbits/s full duplex. Detailed Benchmarking Tests and Results 3 Performance Numbers Under Red Hat Linux 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 TEST #2: VALIDATION OF THROUGHPUT AND CPU UTILIZATION The second test used the same test bench as Test #1 (see Figure 1). The purpose of this test was to verify the throughput measurements of Test #1 with a second independent application. The only limitation of this test is that no CPU utilization measurements are provided. TTCP gives a throughput measurement by sending 16 megabytes (16, 777, 216 bytes) in packets of 8 kilobytes (8192 bytes) to the server. It then divides the total transfer size by the time it took and converts this result to Mbits/s. The throughput obtained, summarized in Table 3, is similar to that obtained from Test #1. The major difference between Test #1 and Test #2 occurs in the method in which the table results were computed. The results from Test #1 are averages computed by the utility itself after automatically performing the test many times. TTCP, however, only performs one data transfer and then displays the computed result. In order to get a reliable value, ten such tests were performed on each port and the arithmetic mean of those ten results was computed. Table 3 displays only the average results that were computed. TA B LE 3: VALIDATION OF THROUGHPUT - TTCP BENCHMARK RESULTS Port 1 (Mbits/s) 10Mbits/s Half duplex 10Mbits/s Full duplex 100Mbits/s Half duplex 100Mbits/s Full duplex 6. 49 8. 94 62. 34 89. 24 Port 2 (Mbits/s) 6. 51 8. 94 61. 56 89. 14 Port 3 (Mbits/s) 6. 54 8. 94 62. 29 89. 26 Port 4 (Mbits/s) 6. 53 8. 94 62. 05 88. 51 At 100Mbits/s full duplex, the results were identical at 89Mbits/s in both Test #1 (as shown in Table 2) and Test #2 (as shown in Table 3, above). At 100Mbits/s half duplex, the results were similar in both Test #1 (as shown in Table 2 at 66Mbits/s) and Test #2 (as shown in Table 3, above at 62Mbits/s). The difference was that the receiving part of the TTCP utility was not run on the server but on a client. In other words, IP forwarding was added to this test, as shown in Figure 3, below. Figure 3: ADDING IP FORWARDING - BENCH CONFIGURATION TCCP Application Client 1 Client 2 Client 3 Client 4 1 IP Forwarding 2 3 4 NS-FNIC/4 Ports in the Server This test verified whether the Matrox NS-FNIC/4 gives similar throughput when used in a router configuration. The reported throughput was similar to those reported by the first three tests and is shown in Table 5. This table only summarizes the results for a packet size of 8KB and a file size of 10MB. [. . . ] The overall value ranges from 20% to 55% with an average of 45%. · Detailed Benchmarking Tests and Results 9 Performance Numbers Under Red Hat Linux 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 TEST #7: COMBINED THROUGHPUT TESTING Test #7 was executed to obtain the combined throughput. It was similar to Test #6 but switches were used to put two clients on each subnet. The final setup for this test was as shown in Figure 3. Figure 5: COMBINED THROUGHPUT TESTING - COMBINED THROUGHPUT SETUP Linux PC with a 10/100 Ethernet card Switch Linux PC with a 10/100 Ethernet card Linux PC with a 10/100 Ethernet card Switch Linux PC with a 10/100 Ethernet card Switch Linux PC with a 10/100 Ethernet card Switch Linux PC with a 10/100 Ethernet card Linux PC with a 10/100 Ethernet card Linux PC with a 10/100 Ethernet card Linux PC with a Matrox NS-FNIC/4 The four available 166MHz Pentiums with Linux installed were used with four additional 100MHz Pentiums to provide the needed eight clients. In order not to corrupt the data obtained, there was one 166MHz Pentium and one 100MHz Pentium on each subnet. [. . . ]