Hello all:
I am working on a project where we have an HP N-class connected to an EMC frame through redundant 1Gb fibre (no load balancing across the two fibre paths). The EMC is configured using metavolumes with 24 disks per meta, so spindle speed should not be an issue.
We are seeing some very erratic performance, most notably, limitations of single threads. For example, if we perform a simple 'dd' on any file on an EMC filesystem we see no more than a max data transfer rate of 21MB/sec. If we try the same command on a a file located on an internal drive we can see 30-35MB/sec, reaching maximum spindle speed.
The best we can ever seem to achieve down a fibre channel is about 50MB/sec.
The ASP vendor cannot (or will not) recognize that 21MB/sec is low and they go so far as to state that HP and EMC say this is adequate. I am at a loss for answers.
Does anyone else have a similar configuration and if so can they see this same type of activity?
Any thoughts would be appreciated.
Regards,
I am working on a project where we have an HP N-class connected to an EMC frame through redundant 1Gb fibre (no load balancing across the two fibre paths). The EMC is configured using metavolumes with 24 disks per meta, so spindle speed should not be an issue.
We are seeing some very erratic performance, most notably, limitations of single threads. For example, if we perform a simple 'dd' on any file on an EMC filesystem we see no more than a max data transfer rate of 21MB/sec. If we try the same command on a a file located on an internal drive we can see 30-35MB/sec, reaching maximum spindle speed.
The best we can ever seem to achieve down a fibre channel is about 50MB/sec.
The ASP vendor cannot (or will not) recognize that 21MB/sec is low and they go so far as to state that HP and EMC say this is adequate. I am at a loss for answers.
Does anyone else have a similar configuration and if so can they see this same type of activity?
Any thoughts would be appreciated.
Regards,