1 |
I think you need to try running a real benchmark like bonnie++ against both. |
2 |
|
3 |
For example, you run "time dd" but you don't include the "sync" in the |
4 |
time... |
5 |
|
6 |
Daniel Iliev wrote: |
7 |
|
8 |
>Neil Bothwick wrote: |
9 |
> |
10 |
> |
11 |
>>Hello Daniel Iliev, |
12 |
>> |
13 |
>> |
14 |
>> |
15 |
>> |
16 |
>>>Actually I'd be glad to read some results from a "Fake RAID-0 vs LVM" |
17 |
>>>tests. My bet would be that RAID-0 w/o LVM would give the best speeds |
18 |
>>> |
19 |
>>> |
20 |
>>> |
21 |
>>Omitting LVM isn't an option, I'd lose all the flexibility that LVM |
22 |
>>offers. I don't see why RAID-0 should be necessarily more efficient than |
23 |
>>LVM, unless there's something superior about RAID-0's striping |
24 |
>>algorithms. I could do some before and after tests, but I'd first have the |
25 |
>>reformat the filesystems to remove any effects of fragmentation. |
26 |
>> |
27 |
>>If no one comes up with a good reason for keeping the RAID, I'll get rid |
28 |
>>of it, running bonnie++ before and after. |
29 |
>> |
30 |
>> |
31 |
>> |
32 |
>> |
33 |
>> |
34 |
> |
35 |
>Hi, Neil! |
36 |
> |
37 |
>Out of curiosity I made some tests which confirmed my expectations. What |
38 |
>about you - did you have time (and wish) to take some performance |
39 |
>benchmarks? I would be glad to see some additional results. |
40 |
> |
41 |
>I'm attaching my tests in file called "bench.txt". |
42 |
> |
43 |
> |
44 |
> |
45 |
> |
46 |
>------------------------------------------------------------------------ |
47 |
> |
48 |
>echo y | mdadm -C /dev/md9 -n2 /dev/sda11 /dev/sdb11 -l0 |
49 |
>mkfs.xfs /dev/md9 |
50 |
>mkdir /test |
51 |
>mount /dev/md9 /test |
52 |
>dd if=/dev/urandom of=/test.rnd bs=1M count=1500 |
53 |
> |
54 |
>time cp /test.rnd /test |
55 |
>real 0m44.981s |
56 |
>user 0m0.036s |
57 |
>sys 0m6.967s |
58 |
> |
59 |
>sync |
60 |
> |
61 |
>time mv /test.rnd /test |
62 |
>real 0m47.514s |
63 |
>user 0m0.047s |
64 |
>sys 0m7.077s |
65 |
> |
66 |
>sync |
67 |
> |
68 |
>time mv /test/test.rnd / |
69 |
>real 0m53.863s |
70 |
>user 0m0.060s |
71 |
>sys 0m8.885s |
72 |
> |
73 |
>mdadm --stop /dev/md9 |
74 |
>pvcreate /dev/sda11 |
75 |
>pvcreate /dev/sdb11 |
76 |
>vgcreate test /dev/sda11 |
77 |
>vgextend test /dev/sdb11 |
78 |
>vgdisplay | grep 'Total PE' |
79 |
> Total PE 1686 |
80 |
>lvcreate -i2 -l1686 -nlogvol test |
81 |
>mkfs.xfs /dev/test/logvol |
82 |
>mount /dev/test/logvol /test |
83 |
> |
84 |
> |
85 |
>time cp /test.rnd /test |
86 |
> |
87 |
>real 1m12.183s |
88 |
>user 0m0.039s |
89 |
>sys 0m9.570s |
90 |
> |
91 |
>sync |
92 |
> |
93 |
>time mv /test.rnd /test |
94 |
> |
95 |
>real 0m51.643s |
96 |
>user 0m0.044s |
97 |
>sys 0m7.275s |
98 |
> |
99 |
>sync |
100 |
> |
101 |
>time mv /test/test.rnd / |
102 |
> |
103 |
>real 1m54.937s |
104 |
>user 0m0.047s |
105 |
>sys 0m9.556s |
106 |
> |
107 |
> |
108 |
>================= |
109 |
>BOTTOM LINE: |
110 |
> |
111 |
>cp /test.rnd /test |
112 |
>LVM: 20.78 [MB/s] |
113 |
>RAID-0: 33.41 [MB/s] |
114 |
> |
115 |
>mv /test.rnd /test |
116 |
>LVM: 29.04[MB/s] |
117 |
>RAID-0: 31.56[MB/s] |
118 |
> |
119 |
>mv /test/test.rnd / |
120 |
>LVM: 11.11[MB/s] |
121 |
>RAID-0: 27.84[MB/s] |
122 |
> |
123 |
>Strange: I repeated the last LVM test because it seemed to me as a low performance peak, but the result was again very low: |
124 |
>time mv /test/test.rnd / |
125 |
> |
126 |
>real 1m27.775s |
127 |
>user 0m0.050s |
128 |
>sys 0m9.813s |
129 |
> |
130 |
>which is: 1500/87.775 = 17.089 [MB/s] |
131 |
> |
132 |
-- |
133 |
gentoo-user@g.o mailing list |