Hi All,
Here is the summary of LMBench(patch2) performance tool results for
different Linux kernels.
There are some surprising results for 2.5.XX kernel.
There is a drastic change in the latency of system calls stat and
open/close between 2.5.xx kernels.
Local communication latencies have also increased for 2.5.44 kernel
compared to 2.5.43.
TCP and AF_UNIX socket stream bandwidth has also got drastic change.
Does any one have any idea why there is a spike in performance of 2.5.xx
kernel ??
Thanks,
pavan
========================================================================
L M B E N C H 2 . 0 [PATCH 2] S U M M A R Y
------------------------------------------------------------------------
----Hardware Specifications: --------------------------------------
# CPU's = 1 - Intel PIII Motherboard = Intel 810 CPU Freq = 868 MHz RAM = 128MB L1 I Cache = 16K L1 D Cache = 16K L2 Cache = 256K File System = ext3 glibc ver. = 2.2.5-34 eth0 = RTL8139
Processor, Processes - times in microseconds - smaller is better ------------------------------------------------------------------------ ------------------------------------ OS MHz null null open selct sig sig fork exec sh call I/O stat close TCP inst hndl proc proc proc ------------------------------- ----- -------- ------ ------- -------- -------- -------- -------- ------------- Linux 2.4.19 868 0.39 0.71 3.90 5.15 29.7 0.94 3.03 143. 734. 5051 Lin 2.4.20-pre11 868 0.39 0.70 3.84 5.08 32.4 0.94 3.03 154. 734. 5059
Linux 2.5.42 868 0.39 0.76 24.6 26.1 32.1 1.01 4.89 361. 1592 7710 Linux 2.5.43 868 0.39 0.75 4.38 5.53 28.9 1.02 3.32 348. 1284 6335 Linux 2.5.44 868 0.39 0.77 24.5 26.1 32.5 1.02 4.89 433. 1587 7634 ------------------------------------------------------------------------ ------------------------------------
Context switching - times in microseconds - smaller is better ------------------------------------------------------------------------ ------------------------------------ OS 2p/0K 2p/16K 2p/64K 8p/16K 8p/64K 16p/16K 16p/64K ctxsw ctxsw ctxsw ctxsw ctxsw ctxsw ctxsw ------------------------------ ----------- ----------- ------------ --------------- ----------- --------------- Linux 2.4.19 1.100 4.3700 13.2 7.3000 181.1 47.6 182.6 Lin 2.4.20-pre11 1.090 4.4000 13.2 9.9000 182.3 46.1 182.9
Linux 2.5.42 1.170 4.5100 13.2 7.4300 187.4 44.7 189.3 Linux 2.5.43 1.250 4.3000 13.2 6.9600 187.9 43.3 188.2 Linux 2.5.44 1.140 4.4000 13.2 7.0200 188.5 43.0 189.2 ------------------------------------------------------------------------ -------------------------------------
*Local* Communication latencies in microseconds - smaller is better ------------------------------------------------------------------------ ------------------------------------ OS 2p/0K Pipe AF UDP RPC/ TCP RPC/ TCP ctxsw UNIX UDP TCP conn ------------------------------- ---------- ---------- ---------- ---------- ----------- --------- ------------ Linux 2.4.19 1.120 6.236 11.2 16.5 37.3 24.1 49.9 83.2 Lin 2.4.20-pre11 1.090 6.426 11.3 18.6 39.7 25.3 52.1 86.1
Linux 2.5.42 1.170 6.913 19.2 28.1 52.4 113.6 145.7 146. Linux 2.5.43 1.250 6.876 12.9 20.7 42.3 29.5 56.2 94.1 Linux 2.5.44 1.140 6.928 19.6 31.1 55.3 113.7 145.4 142. ------------------------------------------------------------------------ -------------------------------------
File & VM system latencies in microseconds - smaller is better ------------------------------------------------------------------------ ------------------------------------ OS 0K File 10K File Mmap Prot Page Create Delete Create Delete Latency Fault Fault ------------------------------- ---------------------- ----------------------- ------------- -------------- Linux 2.4.19 70.8 23.6 235.4 48.4 274.0 0.780 3.00000 Lin 2.4.20-pre11 71.2 25.3 239.8 49.9 269.0 0.784 3.00000
Linux 2.5.42 98.4 52.6 298.4 87.9 511.0 0.972 4.00000 Linux 2.5.43 72.7 26.4 249.1 56.0 505.0 0.917 4.00000 Linux 2.5.44 98.5 53.4 295.2 88.3 503.0 0.956 4.00000 ------------------------------------------------------------------------ -----------------------------------
*Local* Communication bandwidths in MB/s - bigger is better ------------------------------------------------------------------------ ------------------------------------------------ OS Pipe AF TCP File Mmap Bcopy Bcopy Mem Mem UNIX reread reread (libc) (hand) read write ------------------------------- ------- ----------- -------- -------------- ------------- ----------- ----------- --------- Linux 2.4.19 715. 604. 152. 317.6 352.0 111.6 102.6 352. 141.4 Lin 2.4.20-pre11 725. 449. 147. 315.7 352.1 111.6 102.5 351. 141.4
Linux 2.5.42 475. 148. 17.4 307.2 344.5 110.8 102.5 344. 139.5 Linux 2.5.43 387. 617. 38.2 298.1 345.4 111.2 102.4 345. 139.7 Linux 2.5.44 442. 147. 17.0 300.1 344.9 111.0 102.4 345. 139.6 ------------------------------------------------------------------------ --------------------------------------------------
Memory latencies in nanoseconds - smaller is better (WARNING - may not be correct, check graphs) ------------------------------------------------------------------------ ----------------- OS Mhz L1 $ L2 $ Main mem Guesses ----------------------------- -------- ----------- ------------- ------------------------- Linux 2.4.19 868 3.456 8.0640 178.1 Lin 2.4.20-pre11 868 3.456 8.0690 178.2
Linux 2.5.44 868 3.490 8.1560 181.2 Linux 2.5.43 868 3.491 8.1530 180.6 Linux 2.5.44 868 3.490 8.2170 180.4 ------------------------------------------------------------------------ ----------------
Note: All results are average of five iterations.
============================================
============================================ PAVAN KUMAR REDDY N.S. Sr.Software Engineer Wipro Technologies 53/1, Hosur road, Madivala Bangalore - 68. Phone Off: +91-80-5502001-8 extn: 5086. Res: +91-80-6685179 http://www.wipro.com/linux/ ============================================
This archive was generated by hypermail 2b29 : Thu Oct 31 2002 - 22:00:27 EST