All;
Let me appoligize in advance; I'm a UNIX / Linux Admin, so I may not have the big iron terminology correct. Can anyone help us understand an interesting, yet seemingly bizarre and unexpected change in I/O throughput? Any insight in the Mainframe's I/O subsystem interaction with Linux would be greatly appreciated. We have RH v 7.2 running in 3 instances, on a z/800. We have an I/O intensive application seemed to bottleneck at the disk I/O level. To tune the I/O for greater throughput we more than doubled the number CHPIDs from 2 CHIPDs to 5 CHPIDs for VM. As illustrated in the sar output, in almost all measures, I/O seemingly decreased, rather than the expected increase. 08:20:02 AM tps rtps wtps bread/s bwrtn/s .... 10:00:02 AM 1631.95 1152.46 479.49 49424.93 4546.82 10:10:04 AM 1432.91 989.88 443.04 57652.88 3751.61 10:20:03 AM 1594.48 1096.06 498.43 63872.01 4187.39 10:30:02 AM 1561.13 1062.55 498.58 60807.45 4180.14 10:40:04 AM 1591.40 1073.45 517.95 61082.71 4326.86 10:50:02 AM 1546.10 1036.86 509.24 58992.35 4236.89 11:00:04 AM 1542.87 1035.01 507.87 57994.82 4222.20 11:10:03 AM 1523.59 1018.91 504.69 57183.95 4192.64 11:20:08 AM 1585.66 1061.96 523.69 59102.92 4347.70 .... point of CHPID additions ... 11:30:05 AM 289.59 206.67 82.92 10880.54 687.82 11:40:04 AM 320.85 221.71 99.14 11700.20 820.24 11:50:04 AM 323.05 228.10 94.95 11993.89 785.64 12:00:04 PM 350.41 243.22 107.19 13159.21 886.15 12:10:03 PM 244.28 168.34 75.93 9027.88 629.93 12:20:06 PM 354.44 245.26 109.17 13269.73 904.56 12:30:07 PM 318.21 226.82 91.38 11923.86 758.45 12:40:03 PM 330.88 227.47 103.40 12165.62 855.86 12:50:04 PM 322.11 231.72 90.39 12278.75 750.53 01:00:04 PM 335.75 231.46 104.29 12379.37 861.43 Cheers; E! ----------------- Eric Wilson IBM Certified Advanced Technical Expert RedHat Certified Engineer .~. /V\ /( )\ ^^-^^
