Software and hardware annotations 2009 January

This document contains only my personal opinions and calls of judgement, and where any comment is made as to the quality of anybody's work, the comment is an opinion, in my judgement.

[file this blog page at: digg del.icio.us Technorati]

090131 Sat Impressive JFS and eSATA performance
While updating a backup disk using
rsync -v -axHS --delete --no-whole-file /fs/a/ /fs/b/
I was running dstat which reported (note that the disk transfer rate reported is twice the actual transfer rate):
----total-cpu-usage---- -dsk/total- -net/total- ---paging-- ---system--
usr sys idl wai hiq siq| read  writ| recv  send|  in   out | int   csw 
 24  35   0  41   0   0|  63M   62M|1998B 1505B|   0     0 |1252  4270 
 29  38   0  31   1   1|  63M   67M|1938B 1446B|   0     0 |1244  4326 
 11  15   0  73   0   1|  26M   63M|2313B 1650B|   0     0 | 660  1888 
 26  48   0  24   1   1|  75M   62M|1938B 1446B|   0     0 |1435  5042 
 29  43   0  25   1   2|  73M   66M|  11k   28k|   0     0 |1434  4998 
 31  43   0  24   1   1|  70M   70M|  11k 6618B|   0     0 |1378  4764 
 28  51   0  20   0   1|  70M   64M| 105k  9.9k|   0     0 |1497  4898 
 30  36   0  26   4   4|  67M   70M| 101k   42k|   0     0 |1736  5217 
 25  47   0  24   1   3|  71M   66M|  28k 3112B|   0     0 |1423  4874 
 28  49   0  20   1   2|  72M   71M|4252B 4952B|   0     0 |1424  4885 
 27  46   0  24   2   1|  73M   66M|3078B 3286B|   0     0 |1422  4937 
 30  42   0  27   0   1|  71M   71M|3440B 6034B|   0     0 |1515  4856 
 28  48   0  23   1   0|  73M   72M|2283B 1609B|   0     0 |1407  4895 
 24  42   0  32   0   2|  73M   70M|4270B 2044B|   0     0 |1402  4864 
 36  36   0  26   2   0|  72M   70M|2374B 1974B|   0     0 |1381  4821 
 40  36   0  24   0   0|  68M   68M|2012B 1518B|   0     0 |1331  4613 
 28  38   0  31   0   3|  66M   66M|1968B 1463B|   0     0 |1293  4514 
 27  34   0  38   0   1|  62M   62M|2841B 1716B|   0     0 |1229  4203 
 22  50   0  26   1   1|  66M   63M|1860B 1439B|   0     0 |1307  4410 
 31  25   0  42   1   1|  57M   60M|2526B 2070B|   0     0 |1142  3897 
 28  50   0  21   0   1|  73M   71M|1946B 1518B|   0     0 |1417  4879 
 28  49   0  22   1   0|  67M   67M|2200B 1584B|   0     0 |1305  4509 
 32  35   0  32   1   0|  69M   68M|1963B 1380B|   0     0 |1322  4596 
 27  50   0  23   0   0|  72M   73M|1800B 1380B|   0     0 |1393  4808 
 22  45   0  32   0   1|  66M   65M|2314B 1932B|   0     0 |1307  4493 
 25  43   0  32   0   0|  59M   59M|2012B 1518B|   0     0 |1163  4033 
 10  23   0  67   0   0|  27M   57M|2012B 1518B|   0     0 | 644  2021 
 25  48   0  26   0   1|  66M   55M|1963B 1380B|   0     0 |1311  4429 
 34  39   0  26   0   1|  73M   71M|1800B 1380B|   0     0 |1410  4893 
 32  44   0  21   1   2|  73M   80M|2314B 1932B|   0     0 |1438  4931 
 38  43   0  19   0   0|  73M   77M|2150B 1584B|   0     0 |1411  4910 
 26  48   0  24   1   1|  73M   76M|2128B 1643B|   0     0 |1418  4939 
 28  45   0  24   2   1|  72M   79M|2379B 1584B|   0     0 |1399  4820 
 32  44   0  23   1   0|  65M   68M|  19k   28k|   0     0 |1506  4681 
 37  47   0  13   1   2|  70M   71M|5582B 9936B|   0     0 |1687  4963 
 30  44   0  25   0   1|  71M   67M|2788B 3654B|   0     0 |1459  4798 
This was while updating some directories containing mostly large files (large photographs, RPM files), and it was quite impressive. The partitions involved were on the inner cylinders of a 500GB drive and could do about 50MB/s and the reported average of around 35MB/s is pretty good considering it is through the filesystem doing simultaneous reading and writing (obviously not on the same drive).
I get of course much faster rate doing partition image copies, starting with 65MB/s on partitions on the outer cylinders, but the partition synchronized above contained a large filesystems with relatively few changes.