Hi list!
Currently we have 3 DS14mk4 shelves full of 300GB 3,5" 15K FC drives. All drives are in a single aggregate in three raid groups with two spare disks. Time has come to upgrade capacity in the aggregate and i'm thinking of 3 possible ways: [a] buy another DS14mk4 shelf and continue with 300GB drives - safest way, but most expensive per GB [b] buy another DS14mk4 but use 600GB drives - probably good idea, not so expensive as [a] [c] buy DS2246 with 600GB SAS drives - Probably not a good idea, since drives are different type(in this case 2,5" 10K SAS), but it seems that future will be SAS', it's cheaper per GB and takes less space in rack. As i understand, it is possible to use different drive types in the single aggregate(allthough not recommended), but probably will have some impact on performance.
Right now my first choice would be [b] followed by [a]. If anyone would share his/her thoughts on the matter I would be most grateful.
Thanks, Joosep
There is another option, the DS4243 shelf with 15K disks. this can be 300GB / 450GB / 600GB. It is possible to mix RPM's in an aggregate / raidgroup but this takes down the performance to the lowest RPM, so if you would expand the aggregate I would use the same size and RPM.
The cost per GB is one factor to count, but don't forget the performance impact of bigger disks. You can put twice the amount of data on a 600GB disk, but have the same amount of IOPS as a 300GB disk...
Mvg, Wouter Vervloesem
Neoria - Uptime Group Business Park King Square Veldkant 35D B-2550 Kontich
Tel: +32 (0)3 451 23 82 Mailto: wouter.vervloesem@neoria.be Web: http://www.uptimegroup.be
Op 4-nov.-2011, om 15:07 heeft Joosep Meus het volgende geschreven:
Hi list!
Currently we have 3 DS14mk4 shelves full of 300GB 3,5" 15K FC drives. All drives are in a single aggregate in three raid groups with two spare disks. Time has come to upgrade capacity in the aggregate and i'm thinking of 3 possible ways: [a] buy another DS14mk4 shelf and continue with 300GB drives - safest way, but most expensive per GB [b] buy another DS14mk4 but use 600GB drives - probably good idea, not so expensive as [a] [c] buy DS2246 with 600GB SAS drives - Probably not a good idea, since drives are different type(in this case 2,5" 10K SAS), but it seems that future will be SAS', it's cheaper per GB and takes less space in rack. As i understand, it is possible to use different drive types in the single aggregate(allthough not recommended), but probably will have some impact on performance.
Right now my first choice would be [b] followed by [a]. If anyone would share his/her thoughts on the matter I would be most grateful.
Thanks, Joosep _______________________________________________ Toasters mailing list Toasters@teaparty.net http://www.teaparty.net/mailman/listinfo/toasters
This can lead to a sad trail of tears..depending.
Your new added larger drives will only participate in the full aggregate data stripe to the point they are the 'same size' as the rest of the drives in the aggregate.
Data higher than that 'water mark' of 300g deep on the 600g drives, will go to what in effect will perform like a new aggregate of separate 300g drives..you wont have an aggregate wide stripe anymore above that point.
For all of the other problems it can create, just make it a second aggregate by itself.
Theres no big party foul in mixing speeds, most often, people are adding faster ones to slower ones these days, so the change is a net positive.
If you have a massively sensitive IOPS aware application, then dont mix speeds DOWN any.
On Fri, Nov 4, 2011 at 7:23 AM, Vervloesem Wouter < wouter.vervloesem@neoria.be> wrote:
There is another option, the DS4243 shelf with 15K disks. this can be 300GB / 450GB / 600GB. It is possible to mix RPM's in an aggregate / raidgroup but this takes down the performance to the lowest RPM, so if you would expand the aggregate I would use the same size and RPM.
The cost per GB is one factor to count, but don't forget the performance impact of bigger disks. You can put twice the amount of data on a 600GB disk, but have the same amount of IOPS as a 300GB disk...
Mvg, Wouter Vervloesem
Neoria - Uptime Group Business Park King Square Veldkant 35D B-2550 Kontich
Tel: +32 (0)3 451 23 82 Mailto: wouter.vervloesem@neoria.be Web: http://www.uptimegroup.be
Op 4-nov.-2011, om 15:07 heeft Joosep Meus het volgende geschreven:
Hi list!
Currently we have 3 DS14mk4 shelves full of 300GB 3,5" 15K FC drives.
All drives are in a single aggregate in three raid groups with two spare disks.
Time has come to upgrade capacity in the aggregate and i'm thinking of 3
possible ways:
[a] buy another DS14mk4 shelf and continue with 300GB drives - safest
way, but most expensive per GB
[b] buy another DS14mk4 but use 600GB drives - probably good idea, not
so expensive as [a]
[c] buy DS2246 with 600GB SAS drives - Probably not a good idea, since
drives are different type(in this case 2,5" 10K SAS), but it seems that future will be SAS', it's cheaper per GB and takes less space in rack.
As i understand, it is possible to use different drive types in the
single aggregate(allthough not recommended), but probably will have some impact on performance.
Right now my first choice would be [b] followed by [a]. If anyone would
share his/her thoughts on the matter I would be most grateful.
Thanks, Joosep _______________________________________________ Toasters mailing list Toasters@teaparty.net http://www.teaparty.net/mailman/listinfo/toasters
Toasters mailing list Toasters@teaparty.net http://www.teaparty.net/mailman/listinfo/toasters
This can lead to a sad trail of tears..depending.
Your new added larger drives will only participate in the full aggregate data stripe to the point they are the 'same size' as the rest of the drives in the aggregate.
I think you mean raid group here. So long as all disks in a RG are the same size, you don't waste any space. You can have a RG of 300G disks and a different RG of 600G disks in the same aggregate and not waste any disk space.
Data higher than that 'water mark' of 300g deep on the 600g drives, will go to what in effect will perform like a new aggregate of separate 300g drives..you wont have an aggregate wide stripe anymore above that point.
For all of the other problems it can create, just make it a second aggregate by itself.
Theres no big party foul in mixing speeds, most often, people are adding faster ones to slower ones these days, so the change is a net positive.
If you have a massively sensitive IOPS aware application, then dont mix speeds DOWN any.
On Fri, Nov 4, 2011 at 7:23 AM, Vervloesem Wouter < wouter.vervloesem@neoria.be> wrote:
There is another option, the DS4243 shelf with 15K disks. this can be 300GB / 450GB / 600GB. It is possible to mix RPM's in an aggregate / raidgroup but this takes down the performance to the lowest RPM, so if you would expand the aggregate I would use the same size and RPM.
The cost per GB is one factor to count, but don't forget the performance impact of bigger disks. You can put twice the amount of data on a 600GB disk, but have the same amount of IOPS as a 300GB disk...
Mvg, Wouter Vervloesem
Neoria - Uptime Group Business Park King Square Veldkant 35D B-2550 Kontich
Tel: +32 (0)3 451 23 82 Mailto: wouter.vervloesem@neoria.be Web: http://www.uptimegroup.be
Op 4-nov.-2011, om 15:07 heeft Joosep Meus het volgende geschreven:
Hi list!
Currently we have 3 DS14mk4 shelves full of 300GB 3,5" 15K FC drives.
All drives are in a single aggregate in three raid groups with two spare disks.
Time has come to upgrade capacity in the aggregate and i'm thinking of 3
possible ways:
[a] buy another DS14mk4 shelf and continue with 300GB drives - safest
way, but most expensive per GB
[b] buy another DS14mk4 but use 600GB drives - probably good idea, not
so expensive as [a]
[c] buy DS2246 with 600GB SAS drives - Probably not a good idea, since
drives are different type(in this case 2,5" 10K SAS), but it seems that future will be SAS', it's cheaper per GB and takes less space in rack.
As i understand, it is possible to use different drive types in the
single aggregate(allthough not recommended), but probably will have some impact on performance.
Right now my first choice would be [b] followed by [a]. If anyone would
share his/her thoughts on the matter I would be most grateful.
Thanks, Joosep _______________________________________________ Toasters mailing list Toasters@teaparty.net http://www.teaparty.net/mailman/listinfo/toasters
Toasters mailing list Toasters@teaparty.net http://www.teaparty.net/mailman/listinfo/toasters
--
Gustatus Similis Pullus
Steve Losen scl@virginia.edu phone: 434-924-0640
University of Virginia ITC Unix Support
You dont waste SPACE but your end up with two different stripe sizes across the raid group.
ONE stripe ends at 300G deep, another much shorter stripe goes from 301G deel tp 600G deep on the new drives, in the new RG.
..two different IO performance levels in the same aggregate.
..and thats just a simplistic view of the issue. It can get complicated as the physical aggregate filesystem ages.
On Fri, Nov 4, 2011 at 12:47 PM, Steve Losen scl@virginia.edu wrote:
This can lead to a sad trail of tears..depending.
Your new added larger drives will only participate in the full aggregate data stripe to the point they are the 'same size' as the rest of the
drives
in the aggregate.
I think you mean raid group here. So long as all disks in a RG are the same size, you don't waste any space. You can have a RG of 300G disks and a different RG of 600G disks in the same aggregate and not waste any disk space.
Data higher than that 'water mark' of 300g deep on the 600g drives, will
go
to what in effect will perform like a new aggregate of separate 300g drives..you wont have an aggregate wide stripe anymore above that point.
For all of the other problems it can create, just make it a second aggregate by itself.
Theres no big party foul in mixing speeds, most often, people are adding faster ones to slower ones these days, so the change is a net positive.
If you have a massively sensitive IOPS aware application, then dont mix speeds DOWN any.
On Fri, Nov 4, 2011 at 7:23 AM, Vervloesem Wouter < wouter.vervloesem@neoria.be> wrote:
There is another option, the DS4243 shelf with 15K disks. this can be 300GB / 450GB / 600GB. It is possible to mix RPM's in an aggregate / raidgroup but this takes down the performance to the lowest RPM, so if you would expand the aggregate I would use the same size and RPM.
The cost per GB is one factor to count, but don't forget the
performance
impact of bigger disks. You can put twice the amount of data on a 600GB disk, but have the same amount of IOPS as a 300GB disk...
Mvg, Wouter Vervloesem
Neoria - Uptime Group Business Park King Square Veldkant 35D B-2550 Kontich
Tel: +32 (0)3 451 23 82 Mailto: wouter.vervloesem@neoria.be Web: http://www.uptimegroup.be
Op 4-nov.-2011, om 15:07 heeft Joosep Meus het volgende geschreven:
Hi list!
Currently we have 3 DS14mk4 shelves full of 300GB 3,5" 15K FC drives.
All drives are in a single aggregate in three raid groups with two
spare
disks.
Time has come to upgrade capacity in the aggregate and i'm thinking
of 3
possible ways:
[a] buy another DS14mk4 shelf and continue with 300GB drives - safest
way, but most expensive per GB
[b] buy another DS14mk4 but use 600GB drives - probably good idea,
not
so expensive as [a]
[c] buy DS2246 with 600GB SAS drives - Probably not a good idea,
since
drives are different type(in this case 2,5" 10K SAS), but it seems that future will be SAS', it's cheaper per GB and takes less space in rack.
As i understand, it is possible to use different drive types in the
single aggregate(allthough not recommended), but probably will have
some
impact on performance.
Right now my first choice would be [b] followed by [a]. If anyone
would
share his/her thoughts on the matter I would be most grateful.
Thanks, Joosep _______________________________________________ Toasters mailing list Toasters@teaparty.net http://www.teaparty.net/mailman/listinfo/toasters
Toasters mailing list Toasters@teaparty.net http://www.teaparty.net/mailman/listinfo/toasters
--
Gustatus Similis Pullus
Steve Losen scl@virginia.edu phone: 434-924-0640
University of Virginia ITC Unix Support
You dont waste SPACE but your end up with two different stripe sizes across the raid group.
ONE stripe ends at 300G deep, another much shorter stripe goes from 301G deel tp 600G deep on the new drives, in the new RG.
..two different IO performance levels in the same aggregate.
..and thats just a simplistic view of the issue. It can get complicated as the physical aggregate filesystem ages.
Sorry Jeff, I didn't read your post carefully enough.
Steve Losen scl@virginia.edu phone: 434-924-0640
University of Virginia ITC Unix Support
-
Jeff Mohler -
Joosep Meus -
Steve Losen -
Vervloesem Wouter