1. Storage Team (Controller Performance) Aggregate Utilization
End User
Application Team
Server Team
Architect Team
Storage Team (Controller Performance)
Aggregate Utilization
Storage Team
Architect Team
End User Complaint
Application Team Investigate
Server Team Investigate
Storage Team Investigate
Open a NetApp Case
Review IMT
Review SLA
Review CPU
& Memory Requirement
Open a Performance case
HBA Firmware
Review Controller Performance
Workload Assessment
Workload Assessment
New
Review Aggr Utilization
NOT OK
Existing
For an SLA of no latency impact during TO Maintain HA combined Node Utilization less than 100%
Read/Write? Sequential vs Random
IO Size ?
NOT OK
Review NetApp FC/ iSCSI HUK,
relevant MS Hotfixes and Linux/Unix Hotfixes
Node Utilization Greater than 85%
What has changed
SAN Switch Firmware
Review System Event Log or /var/log
Review whether we have enough target port.
Target Queue Depth Port is
2048
Review Disk Utilization
Run the perfstat on the host where there is a performance issue with the controller.
You must capture the period during the performance hit.
Thereafter, upload the perfstat to the case owner
Only applies to SAS and SATA only
For an SLA where latency impact during TO is an acceptable risk Maintain HA combined Node Utilization less than 150%
Review Application Event Log
Review Application and Storage Vendor - Best Practises
Controller ONTAP
version
Review MPIO
Is the load balance across all Nodes
Configuration
Aggregate Disk busy greater than 50%
Consider moving
the load to other
aggregate or Node
NOT OK
> 50%
Application still having Performance Issue
Resolve Application Issue
Maintain aggregate utilization less than 50%
Not Balanced
Microsoft MPIO
{Make sure its set to LQD}
NetApp DSM
Review Process
Review Mulitpath on Linux/Unix
Perform a volume
Add more Target Ports
move to another Node No
Open a
No software Vendor Case
Review IOPS
Did Application Improve
Engage Duty Manager if issue is not resolved within satisfaction
The load is balanced
across multiple nodes
Can we clamp the application down with QoS
Case Closed
Calculate the total number of Queue Depth for all the hosts
NOT OK
Yes
Review HBA Queue Depth
Yes
SSD can yield as high as IOP/s
SAS random access time ~7ms
Gives ~120 IOP/s per data spindle
SATA random access time ~10ms
Gives ~80 IOP/s per data spindle
Case Closed
Tune HBA Queue Depth if necessary
Perform a QoS against the LUN or Volume
Review the Performance OK
Review the Performance
Case Closed
IOPS required exceed the number of data spindles. Consider buying more disk spindles or move the load to SSD
Not OK
Performance OK
Case Closed
Open Apps/OS Vendor Case

2. therecsaosleved ownweirthin satisfaction
End User
Application Team
Server Team
Architect Team
Storage Team (Controller Performance)
Storage Team (Aggregate Utilization)
Storage Team
Architect Team
Open a NetApp Case
End User Complaint
Application Team Investigate
Server Team Investigate
Storage Team Investigate
Review IMT
Review SLA
Open a Performance case
UCS
Firmware
Review CPU
& Memory Requirement
Workload Assessment
Workload Assessment
Review Controller Performance
New
For an SLA of no latency impact during TO Maintain HA combined Node Utilization less than 100%
NOT OK
Review Aggr
Utilization
Run the perfstat on the host where there is a performance issue with the controller.
You must capture the period during the performance
hit.
Controller ONTAP
version
Existing
Review System Event Log or
/var/log
Read/Write ?□
What has changed
Sequential vs Random□
For an SLA where latency impact during TO is an acceptable risk
Maintain HA combined Node Utilization less than 150%
Node
Utilization Greater than 85%
IO Size ?
Ensure that
NOT OK OS
(Ensure all OS tuning are in place)
we are not saturating the 10gE link.
Review Disk Utilization
Review
Review Application Event Log
Application and Storage Vendor - Best Practises
Only applies to SAS and SATA only
Is the load balance across all Nodes
Aggregate
Disk busy greater than
50%
Consider moving the load to other aggregate or Node
Maintain aggregate utilization less than 50%
Application still having Performance Issue
Resolve Application Issue
Vmware
Hyper-V
Native OS
> 50%
Not Balanced
Add more
Perform a volume move to another Node
Review Process
10gE Ports to the SVM
Ensure NFS
TEhnegraegaefteDru, ty
Ensure VSC
is properly configured on all ESX Hosts
Ensure all necessary hotfixes are applied
mounts are properly configured based on best
Open a software Vendor Case
upMloaandagtheer if piessrfusetaitstonot
therecsaosleved ownweirthin satisfaction No
Review IOPS
Application Improve NO
Case Closed
The load is balanced
Can we clamp the application down with QoS
practises
across multiple nodes
Ensure
SSD can yield as high as IOP/s
Ensure NFS
YES
Jumbo Frame is enabled from the Hosts to the
mounts are properly configured based on best practises.
Please review Best Practises Guides on NFS
SAS random access time
~7ms
Gives ~120 IOP/s per data spindle
NOT OK
Yes
Controllers
Case Closed
SATA random access time
~10ms
Gives ~80 IOP/s per data
spindle
Perform a QoS against the LUN or Volume
IOPS
required
Review the
Review the Performance
Performance
NOT OK
exceed the number of data spindles.
Consider buying more disk spindles or move the
Not OK
Performance OK
Case Closed
Performance is OK
Open Apps/OS Vendor Case
Case Closed
load to SSD

3. mds9000/sw/rel_3_x/troubleshooting/guide/ts_guide.pdf