1
# from http://www.miracleas.com/BAARF/RAID5_versus_RAID10.txt
2
# also see http://www.miracleas.com/BAARF/BAARF2.html
4
# Note: I, the Debian maintainer, do not agree with some of the arguments,
5
# especially not with the total condemning of RAID5. Anyone who talks about
6
# data loss and blames the RAID system should spend time reading up on Backups
7
# instead of trying to evangelise, but that's only my opinion. RAID5 has its
8
# merits and its shortcomings, just like any other method. However, the author
9
# of this argument puts forth a good case and thus I am including the
10
# document. Remember that you're the only one that can decide which RAID level
14
RAID5 versus RAID10 (or even RAID3 or RAID4)
16
First let's get on the same page so we're all talking about apples.
20
OK here is the deal, RAID5 uses ONLY ONE parity drive per stripe and many
21
RAID5 arrays are 5 (if your counts are different adjust the calculations
22
appropriately) drives (4 data and 1 parity though it is not a single drive
23
that is holding all of the parity as in RAID 3 & 4 but read on). If you
24
have 10 drives or say 20GB each for 200GB RAID5 will use 20% for parity
25
(assuming you set it up as two 5 drive arrays) so you will have 160GB of
26
storage. Now since RAID10, like mirroring (RAID1), uses 1 (or more) mirror
27
drive for each primary drive you are using 50% for redundancy so to get the
28
same 160GB of storage you will need 8 pairs or 16 - 20GB drives, which is
29
why RAID5 is so popular. This intro is just to put things into
32
RAID5 is physically a stripe set like RAID0 but with data recovery
33
included. RAID5 reserves one disk block out of each stripe block for
34
parity data. The parity block contains an error correction code which can
35
correct any error in the RAID5 block, in effect it is used in combination
36
with the remaining data blocks to recreate any single missing block, gone
37
missing because a drive has failed. The innovation of RAID5 over RAID3 &
38
RAID4 is that the parity is distributed on a round robin basis so that
39
there can be independent reading of different blocks from the several
40
drives. This is why RAID5 became more popular than RAID3 & RAID4 which
41
must sychronously read the same block from all drives together. So, if
42
Drive2 fails blocks 1,2,4,5,6 & 7 are data blocks on this drive and blocks
43
3 and 8 are parity blocks on this drive. So that means that the parity on
44
Drive5 will be used to recreate the data block from Disk2 if block 1 is
45
requested before a new drive replaces Drive2 or during the rebuilding of
46
the new Drive2 replacement. Likewise the parity on Drive1 will be used to
47
repair block 2 and the parity on Drive3 will repair block4, etc. For block
48
2 all the data is safely on the remaining drives but during the rebuilding
49
of Drive2's replacement a new parity block will be calculated from the
50
block 2 data and will be written to Drive 2.
52
Now when a disk block is read from the array the RAID software/firmware
53
calculates which RAID block contains the disk block, which drive the disk
54
block is on and which drive contains the parity block for that RAID block
55
and reads ONLY the one data drive. It returns the data block. If you
56
later modify the data block it recalculates the parity by subtracting the
57
old block and adding in the new version then in two separate operations it
58
writes the data block followed by the new parity block. To do this it must
59
first read the parity block from whichever drive contains the parity for
60
that stripe block and reread the unmodified data for the updated block from
61
the original drive. This read-read-write-write is known as the RAID5 write
62
penalty since these two writes are sequential and synchronous the write
63
system call cannot return until the reread and both writes complete, for
64
safety, so writing to RAID5 is up to 50% slower than RAID0 for an array of
65
the same capacity. (Some software RAID5's avoid the re-read by keeping an
66
unmodified copy of the orginal block in memory.)
70
RAID10 is one of the combinations of RAID1 (mirroring) and RAID0
71
(striping) which are possible. There used to be confusion about what
72
RAID01 or RAID10 meant and different RAID vendors defined them
73
differently. About five years or so ago I proposed the following standard
74
language which seems to have taken hold. When N mirrored pairs are
75
striped together this is called RAID10 because the mirroring (RAID1) is
76
applied before striping (RAID0). The other option is to create two stripe
77
sets and mirror them one to the other, this is known as RAID01 (because
78
the RAID0 is applied first). In either a RAID01 or RAID10 system each and
79
every disk block is completely duplicated on its drive's mirror.
80
Performance-wise both RAID01 and RAID10 are functionally equivalent. The
81
difference comes in during recovery where RAID01 suffers from some of the
82
same problems I will describe affecting RAID5 while RAID10 does not.
84
Now if a drive in the RAID5 array dies, is removed, or is shut off data is
85
returned by reading the blocks from the remaining drives and calculating
86
the missing data using the parity, assuming the defunct drive is not the
87
parity block drive for that RAID block. Note that it takes 4 physical
88
reads to replace the missing disk block (for a 5 drive array) for four out
89
of every five disk blocks leading to a 64% performance degradation until
90
the problem is discovered and a new drive can be mapped in to begin
91
recovery. Performance is degraded further during recovery because all
92
drives are being actively accessed in order to rebuild the replacement
95
If a drive in the RAID10 array dies data is returned from its mirror drive
96
in a single read with only minor (6.25% on average for a 4 pair array as a
97
whole) performance reduction when two non-contiguous blocks are needed from
98
the damaged pair (since the two blocks cannot be read in parallel from both
99
drives) and none otherwise.
101
One begins to get an inkling of what is going on and why I dislike RAID5,
102
but, as they say on late night info-mercials, there's more.
104
What's wrong besides a bit of performance I don't know I'm missing?
106
OK, so that brings us to the final question of the day which is: What is
107
the problem with RAID5? It does recover a failed drive right? So writes
108
are slower, I don't do enough writing to worry about it and the cache
109
helps a lot also, I've got LOTS of cache! The problem is that despite the
110
improved reliability of modern drives and the improved error correction
111
codes on most drives, and even despite the additional 8 bytes of error
112
correction that EMC puts on every Clariion drive disk block (if you are
113
lucky enough to use EMC systems), it is more than a little possible that a
114
drive will become flaky and begin to return garbage. This is known as
115
partial media failure. Now SCSI controllers reserve several hundred disk
116
blocks to be remapped to replace fading sectors with unused ones, but if
117
the drive is going these will not last very long and will run out and SCSI
118
does NOT report correctable errors back to the OS! Therefore you will not
119
know the drive is becoming unstable until it is too late and there are no
120
more replacement sectors and the drive begins to return garbage. [Note
121
that the recently popular IDE/ATA drives do not (TMK) include bad sector
122
remapping in their hardware so garbage is returned that much sooner.]
123
When a drive returns garbage, since RAID5 does not EVER check parity on
124
read (RAID3 & RAID4 do BTW and both perform better for databases than
125
RAID5 to boot) when you write the garbage sector back garbage parity will
126
be calculated and your RAID5 integrity is lost! Similarly if a drive
127
fails and one of the remaining drives is flaky the replacement will be
128
rebuilt with garbage also propagating the problem to two blocks instead of
131
Need more? During recovery, read performance for a RAID5 array is
132
degraded by as much as 80%. Some advanced arrays let you configure the
133
preference more toward recovery or toward performance. However, doing so
134
will increase recovery time and increase the likelihood of losing a second
135
drive in the array before recovery completes resulting in catastrophic
136
data loss. RAID10 on the other hand will only be recovering one drive out
137
of 4 or more pairs with performance ONLY of reads from the recovering pair
138
degraded making the performance hit to the array overall only about 20%!
139
Plus there is no parity calculation time used during recovery - it's a
142
What about that thing about losing a second drive? Well with RAID10 there
143
is no danger unless the one mirror that is recovering also fails and
144
that's 80% or more less likely than that any other drive in a RAID5 array
145
will fail! And since most multiple drive failures are caused by
146
undetected manufacturing defects you can make even this possibility
147
vanishingly small by making sure to mirror every drive with one from a
148
different manufacturer's lot number. ("Oh", you say, "this schenario does
149
not seem likely!" Pooh, we lost 50 drives over two weeks when a batch of
150
200 IBM drives began to fail. IBM discovered that the single lot of
151
drives would have their spindle bearings freeze after so many hours of
152
operation. Fortunately due in part to RAID10 and in part to a herculean
153
effort by DG techs and our own people over 2 weeks no data was lost.
154
HOWEVER, one RAID5 filesystem was a total loss after a second drive failed
155
during recover. Fortunately everything was on tape.
157
Conclusion? For safety and performance favor RAID10 first, RAID3 second,
158
RAID4 third, and RAID5 last! The original reason for the RAID2-5 specs
159
was that the high cost of disks was making RAID1, mirroring, impractical.
160
That is no longer the case! Drives are commodity priced, even the biggest
161
fastest drives are cheaper in absolute dollars than drives were then and
162
cost per MB is a tiny fraction of what it was. Does RAID5 make ANY sense
163
anymore? Obviously I think not.
165
To put things into perspective: If a drive costs $1000US (and most are far
166
less expensive than that) then switching from a 4 pair RAID10 array to a 5
167
drive RAID5 array will save 3 drives or $3000US. What is the cost of
168
overtime, wear and tear on the technicians, DBAs, managers, and customers
169
of even a recovery scare? What is the cost of reduced performance and
170
possibly reduced customer satisfaction? Finally what is the cost of lost
171
business if data is unrecoverable? I maintain that the drives are FAR
172
cheaper! Hence my mantra:
174
NO RAID5! NO RAID5! NO RAID5! NO RAID5! NO RAID5! NO RAID5! NO RAID5!