File:Gmirror-performance.png - Revision history

Jimbo: added Linux data

2007-12-28T02:45:19Z

added Linux data

Jimbo at 04:09, 27 December 2007

2007-12-27T04:09:01Z

Jimbo at 01:58, 27 December 2007

2007-12-27T01:58:08Z

Jimbo at 01:53, 27 December 2007

2007-12-27T01:53:27Z

Jimbo: Chart showing performance of gmirror (and some proprietary RAID1 hardware implementations) under amd64 6.2-RELEASE, Athlon X2 5000+, 2GB RAM. Each file copied in the graph was individually generated using dd if=/dev/random bs=16m count=200 (3200MB file).

2007-12-27T01:48:47Z

Chart showing performance of gmirror (and some proprietary RAID1 hardware implementations) under amd64 6.2-RELEASE, Athlon X2 5000+, 2GB RAM. Each file copied in the graph was individually generated using dd if=/dev/random bs=16m count=200 (3200MB file).

New page

@@ Line 2: / Line 2: @@
   '''System: '''
-     amd64 6.2-RELEASE, Athlon X2 5000+, 2GB RAM
+     FreeBSD OS: amd64 6.2-RELEASE
 x Western Digital 250GB drives (WDC WD2500JS-22NCB1 10.02E02 SATA-300)
 x Western Digital 500GB drives (WDC WD5000AAKS-00YGA0 12.01C02)
-     Nvidia nForce onboard RAID controller, Promise TX2300 RAID controller
+     Nvidia nForce onboard RAID controller
 Each process shown in the graph is a cp to /dev/null of an individually generated 3200MB random binary file
@@ Line 12: / Line 16: @@
 The "split" balance algorithm was also tested but resolved in universally horrible performance, so was not included on the graph.
-For these tests, read-ahead cache value was GREATLY increased from 6.2-RELEASE's default of 8: '''sysctl -w vfs.read_max=128'''.
+For these tests, FreeBSD's read-ahead cache value was GREATLY increased from 6.2-RELEASE's default of 8: '''sysctl -w vfs.read_max=128'''.  Similarly, Ubuntu's read-ahead cache value was drastically increased from system default of '''hdparm /dev/md0 -a256''' to '''hdparm /dev/md0 -a4096'''.  Changing these values made TREMENDOUS differences in throughput - frequently doubling, sometimes even quintupling average transfer rate across tests.
 Raw data follows:
-  Nvidia onboard RAID1: 2x250		60	108	90	96	100	* 15 secs delta on 2-process copy / 38 seconds on 5-process copy!
+ Linux RAID1: 3x250/2x500		68	152	213	251	258	* Hdparm -a 4096
   Gmirror: 3x250/2x500 round-robin	63	73	69	71	72
   Baseline: 1x500			76	70	72	72	72

← Older revision		Revision as of 04:09, 27 December 2007
Line 11:		Line 11:

	The "split" balance algorithm was also tested but resolved in universally horrible performance, so was not included on the graph.		The "split" balance algorithm was also tested but resolved in universally horrible performance, so was not included on the graph.
		+
		+	For these tests, read-ahead cache value was GREATLY increased from 6.2-RELEASE's default of 8: '''sysctl -w vfs.read_max=128'''.

	Raw data follows:		Raw data follows:

← Older revision		Revision as of 01:58, 27 December 2007
Line 9:		Line 9:
	Each process shown in the graph is a cp to /dev/null of an individually generated 3200MB random binary file		Each process shown in the graph is a cp to /dev/null of an individually generated 3200MB random binary file
	(using dd if=/dev/random bs=16m count=200).		(using dd if=/dev/random bs=16m count=200).
		+
		+	The "split" balance algorithm was also tested but resolved in universally horrible performance, so was not included on the graph.
		+
		+	Raw data follows:
		+
		+	Nvidia onboard RAID1: 2x250 60 108 90 96 100 * 15 secs delta on 2-process copy / 38 seconds on 5-process copy!
		+	Gmirror: 3x250/2x500 round-robin 63 73 69 71 72
		+	Baseline: 1x500 76 70 72 72 72
		+	Baseline: 1x250 56 56 55 55 56
		+	Gmirror: 3x250 round-robin 59 53 53 56 55
		+	Promise TX-2300 RAID1: 2x250 61 52 54 55 57 * 15 sec delta on 5-process copy
		+	Gmirror: 3x250/2x500 split 128k 54 50 47 47 47
		+	Gmirror: 2x250 round-robin 58 49 51 51 50
		+	Gmirror: 2x250 split 128k 55 47 47 47 48
		+	Gmirror: 3x250 split 128k 55 46 49 55 54
		+
		+	"Delta" in raw numbers refers to individual processes finishing significantly faster than others, though all were started simultaneously.

← Older revision		Revision as of 01:53, 27 December 2007
Line 1:		Line 1:
−	Chart showing performance of gmirror (and some proprietary RAID1 hardware implementations) ~~under~~ amd64 6.2-RELEASE, Athlon X2 5000+, 2GB RAM. Each ~~file copied~~ in the graph ~~was~~ individually generated using dd if=/dev/random bs=16m count=200 ~~(3200MB file~~).	+	Chart showing performance of gmirror (and some proprietary RAID1 hardware implementations).
		+
		+	'''System: '''
		+	amd64 6.2-RELEASE, Athlon X2 5000+, 2GB RAM
		+	3x Western Digital 250GB drives (WDC WD2500JS-22NCB1 10.02E02 SATA-300)
		+	2x Western Digital 500GB drives (WDC WD5000AAKS-00YGA0 12.01C02)
		+	Nvidia nForce onboard RAID controller, Promise TX2300 RAID controller
		+
		+	Each process shown in the graph is a cp to /dev/null of an individually generated 3200MB random binary file
		+	(using dd if=/dev/random bs=16m count=200).