HP StorageWorks Scalable File Share - Manual (Page 336)

Performance figures B–8
B.5 Default file system stripe count
Figure B-4 Aggregate bandwidth versus file stripe count
The graph in Figure B-4 shows the variation in aggregate file system bandwidth for write operations as the
file stripe count used by the client nodes is varied. The configuration on which these results are based is as
follows:
• 16 Object Storage Servers, with four SFS20 arrays attached to each Object Storage Server (total of
64 SFS20 arrays)
• Each array populated with twelve 250GB disks, configured as one 2TB LUN with ADG (RAID6)
redundancy (that is, a total of 64 OST LUNs, one LUN on each of the SFS20 arrays)
• One file system, using 64 OST services
All results are based on tests using new, empty file systems with default MDS and OST mount options and
without quotas enabled.
The variations were measured for 64, 128, and 192 client nodes. Each client node wrote one 4GB file to
the file system repeatedly, with varying file stripe counts of 1, 4, 8, and 16.
The results show that the aggregate file system bandwidth is highest where the file stripe count is one. This
is the stripe count that is suggested as the default for the file system when the file system is being created.
Note that the bandwidth possible for any one file can be increased by striping the file over more OST
services.
Bandwidth vs Stripe count (64 OSTs)
0
1000
2000
3000
4000
5000
6000
7000
1 4 8 16
Stripe count
MB
/s
ec 64 clients
128 Clients
192 clients