RAID Controller Failover

I'm an IT admin starting to look at building a SAN for my company and I'm
curious if anyone out there can explain how some of the popular RAID vendors
(i.e. EMC, Chaparral, Infortrend) handle failover (resuming I/O with another
RAID controller after one has failed)? I'm mostly interested in failover on
the storage side (as opposed to the host side).

Specifically I'm interested in knowing if failover is generally accomplished
by a surviving controller taking over the failed controller's (or failed
port's) AL_PA('s) or if surviving controllers actually alias failed
controller's WWN's?

Or is this something that's generally handled at the switch level.

I'm trying to better understand how failover is accomplished transparent to
the host. Many thanks for any input regarding this.

i'm not an IT manager

but a raid 5 solution is imho the best!


--
can8

On Fri, 07 Nov 2003 21:33:49 +0000, Steve Holly wrote:

I'm an IT admin starting to look at building a SAN for my company and I'm
curious if anyone out there can explain how some of the popular RAID vendors
(i.e. EMC, Chaparral, Infortrend) handle failover (resuming I/O with another
RAID controller after one has failed)? I'm mostly interested in failover on
the storage side (as opposed to the host side).

Specifically I'm interested in knowing if failover is generally accomplished
by a surviving controller taking over the failed controller's (or failed
port's) AL_PA('s) or if surviving controllers actually alias failed
controller's WWN's?

I'm assuming that you are talking about failover within the same
storage unit, and not between two physical units. I'm not too familar with
how EMC does it, but most (I'm sure not all; everyone does things
different) vendors will present the controllers with one WWN and the
failover is completely transparent to the host.

Some vendors will also distribute the controllers with seperate WWN's
which will rely on the host to fail over. This will be controlled similar
to that of a lost disk/path failover where each controller is its own
path to the same disk. When that path is lost (or the controller dies)
then the software, LVM or vendor software fails the I/O over after a
certain amount of time.

If you are talking about a failover between two physical arrays (this
would only happen under very strange circumstances) then this will have to
be handled by another piece of software. Possibely a high availability
package, or LVM where if the disk along with all paths are lost. So the
software in this case is resposible fo detecting a failure and switching
to the secondary disk.

I hope this helps.

- Jake

If you want to save money on disks - yes.
If the disk drive cost is neglectable for you - then RAID1+0 is better by
far.

RAID4 and RAID5 are very slow on writes.

--
Maxim Shatskih, Windows DDK MVP
StorageCraft Corporation

http://www.storagecraft.com


"canotto" wrote in message
...
i'm not an IT manager

but a raid 5 solution is imho the best!


--
can8

On Sat, 08 Nov 2003 18:45:17 +0300, Maxim S. Shatskih wrote:

If you want to save money on disks - yes.
If the disk drive cost is neglectable for you - then RAID1+0 is better by
far.

RAID4 and RAID5 are very slow on writes.

I've noticed as the controllers get more advanced that the caching and
other alogrithms used minimize the actual write times that the host
system sees. My tests on the recent HP equipment show that the difference
in write times between RAID-1 and RAID-5 are within 1MB of each other. I
have a hard time believing that anyone would be driven away from RAID-5
due to performance factors on high-end equipment. I have the bonnie++
stats if you'd like to see them.

- Jake

Jake Roersma wrote:

I've noticed as the controllers get more advanced that the caching and
other alogrithms used minimize the actual write times that the host
system sees. My tests on the recent HP equipment show that the difference
in write times between RAID-1 and RAID-5 are within 1MB of each other. I
have a hard time believing that anyone would be driven away from RAID-5
due to performance factors on high-end equipment. I have the bonnie++
stats if you'd like to see them.

Random small writes can still kill you. One write turns into
read-read-write-write. Latency doubles, throughput is a quarter.


Thomas

Hi Jake,

I've noticed as the controllers get more advanced that the caching and
other alogrithms used minimize the actual write times that the host
system sees. My tests on the recent HP equipment show that the difference
in write times between RAID-1 and RAID-5 are within 1MB of each other. I
have a hard time believing that anyone would be driven away from RAID-5
due to performance factors on high-end equipment. I have the bonnie++
stats if you'd like to see them.
===

I'd be interested, would you inf sharing/posting them along with the
specs for the hardware you've used?

Many thanks in advance,

Jochen

On Sun, 09 Nov 2003 10:58:51 +0100, Jochen Kaiser wrote:

Hi Jake,

I'd be interested, would you inf sharing/posting them along with the
specs for the hardware you've used?

Many thanks in advance,

Jochen

Jochen,
Here is the link to my testing (sorry due to the amount of information i
had to leave it as html). As time progresses I will be adding more to
it. If you need anymore information on the hardware/software please let
me know. I tried to be as scientific as possible, each test result is the
average of three test results. They were also done before the EVA 5000
was placed into production so there was no other traffic on the controller
and drives. All volumes were created with 80GB of space so they should
be spread out across the same number of disks. RAID-1 is actually RAID-1+0.

http://www.copiosus.net/bonnie/results.html

- Jake

Jake,

Jochen,
Here is the link to my testing (sorry due to the amount of information i
had to leave it as html). As time progresses I will be adding more to
it. If you need anymore information on the hardware/software please let
me know. I tried to be as scientific as possible, each test result is the
average of three test results. They were also done before the EVA 5000
was placed into production so there was no other traffic on the controller
and drives. All volumes were created with 80GB of space so they should
be spread out across the same number of disks. RAID-1 is actually RAID-1+0.

http://www.copiosus.net/bonnie/results.html

Thank you very much. I wouldn't have thought that performance of both
Raid 1+0 and Raid 5 are that close, the same is true for the disks. I
would've expected the 10k disks to be much slower in randon r/w than the
15k disks. Makes me rethink our database sizing concept.

BTW, have you ever tested Raid 1+0 vs. 5 concerning a lot of small r/w
transactions? (Like typical OLTP rdbm system behavior)

Thanks again,

Jochen

"Jochen Kaiser" wrote in message
...
Jake,

Jochen,
Here is the link to my testing (sorry due to the amount of information i
had to leave it as html). As time progresses I will be adding more to
it. If you need anymore information on the hardware/software please let
me know. I tried to be as scientific as possible, each test result is
the
average of three test results. They were also done before the EVA 5000
was placed into production so there was no other traffic on the
controller
and drives. All volumes were created with 80GB of space so they should
be spread out across the same number of disks. RAID-1 is actually
RAID-1+0.

http://www.copiosus.net/bonnie/results.html

Thank you very much. I wouldn't have thought that performance of both
Raid 1+0 and Raid 5 are that close, the same is true for the disks. I
would've expected the 10k disks to be much slower in randon r/w than the
15k disks. Makes me rethink our database sizing concept.

Well, to be precise, the performance of RAID-1 and RAID-5 (and the
performance of disks) varies a lot more than those results indicate: among
other things, the use of extensive stable write-back cache and (on Linux)
reiserfs just smooths out a lot of this variation.

But it can't completely obscure the fundamentals. While per-operation
latency on a lightly-loaded system may seem similar, when the going gets
tough down at the disk level for a given usable storage size RAID-1 offers
nearly twice the streaming sequential read bandwidth and nearly twice the
read IOPS that RAID-5 does. And for *truly* random small write operations
(rather than operations like file creation/deletion where much of the
updating is concentrated in a small number of blocks - parent-directory
updates, for example) where all that the large write-back cache can do is
allow the writes to be queue-optimized rather than coalesced, RAID-1 will
out-perform RAID-5 by a factor of around 2.


BTW, have you ever tested Raid 1+0 vs. 5 concerning a lot of small r/w
transactions? (Like typical OLTP rdbm system behavior)

That might well tend to highlight the differences better (as long as the
database working set significantly exceeded the size of the cache), for the
reasons noted above.

- bill