Using fio with adjusted parameters for use case "Hana DB"

Using fio with adjusted parameters for use case "Hana DB"

[Thomas Schneider]

Hello,

I'm starting evaluating usage of fio <https://github.com/axboe/fio>.

My goal is to benchmark different storage device types, e.g. SSD and 
shared storage, that is close to the I/O workload generated by the 
"real" application: Hana DB.

But before defining this "application related benchmark" I would run a 
generic benchmark first that gives an idea of the IO performance of the 
relevant storage device; this generic benchmark is using options that 
are well known to deliver meaningful results with an acceptable (short) 
runtime.

The idea behind this strategy is:
If the generic benchmark already indicates poor IO performance, there's 
no need to execute another benchmark that simulates the "real" application.

Could you please share some information of this generic benchmark(s) for 
devices SSD, RAID and NFS share (e.g. NetApp, EMC)?


Regards
Thomas

Re: Using fio with adjusted parameters for use case "Hana DB"

[Vincent Fu]

On 3/7/23 09:30, Thomas Schneider wrote:
> Hello,
> 
> I'm starting evaluating usage of fio <https://github.com/axboe/fio>.
> 
> My goal is to benchmark different storage device types, e.g. SSD and 
> shared storage, that is close to the I/O workload generated by the 
> "real" application: Hana DB.
> 
> But before defining this "application related benchmark" I would run a 
> generic benchmark first that gives an idea of the IO performance of the 
> relevant storage device; this generic benchmark is using options that 
> are well known to deliver meaningful results with an acceptable (short) 
> runtime.
> 
> The idea behind this strategy is:
> If the generic benchmark already indicates poor IO performance, there's 
> no need to execute another benchmark that simulates the "real" application.
> 
> Could you please share some information of this generic benchmark(s) for 
> devices SSD, RAID and NFS share (e.g. NetApp, EMC)?
> 
> 
> Regards
> Thomas
> 

Thomas, you could start with this:

https://www.snia.org/tech_activities/standards/curr_standards/pts

Vincent

Using fio with adjusted parameters for use case "Hana DB"

[Thomas Schneider]

Hello Vincent,

I'm starting evaluating usage of fio <https://github.com/axboe/fio>.

My goal is to benchmark different storage device types, e.g. SSD and 
shared storage, that is close to the I/O workload generated by the 
"real" application: Hana DB.

But before defining this "application related benchmark" I would run a 
generic benchmark first that gives an idea of the IO performance of the 
relevant storage device; this generic benchmark is using options that 
are well known to deliver meaningful results with an acceptable (short) 
runtime.

The idea behind this strategy is:
If the generic benchmark already indicates poor IO performance, there's 
no need to execute another benchmark that simulates the "real" application.

Could you please share some information of this generic benchmark(s) for 
devices SSD, RAID and NFS share (e.g. NetApp, EMC)?


Regards
Thomas