Three questions from a customer

[bachman]

I have a customer who is testing Balance 4.1.0. He asks the following three questions:

1. What are the “uncounted for – unmonitored” workloads
2. How do we get a deeper view into the “other” IOPS
3. Is it possible to increase the collection frequency less than 1 hour? (30min, 5min, etc?)

P.

[plauterb · Accepted solution]

What are the “uncounted for – unmonitored” workloads

These are guests or physical servers that are not directly monitored by Balance. It represents front-end IO that is not attributed to a particular server or ESX host. If there is unaccounted IO on the disks that we cannot attribute to an external source, is considered an internal workload.

How do we get a deeper view into the “other” IOPS

"Other" IOPS are operations that are not reads nor writes. You should only see this on NFS and CIFS volumes, not block LUNs. Network protocols like NFS and CIFS have non-data operations that are metadata, such a open directory, find first file, or get file attributes. Balance does not break these down further, but there are other methods to get this detail. Not sure of the method of how to do this.

Is it possible to increase the collection frequency less than 1 hour? (30min, 5min, etc?)

Balance performance statistical collection is every 15 minutes, Balance analysis runs once per hour, using the 15 minute samples. The collection intervals and schedule are not configurable, but are very low impact on the monitored devices.

[mavinash]

Other_ops on a volume are the summation of "all other ops" which are not "read" nor "write". These can be broadly categorized as internal ops and external ops.

They may be metadata operations, directory look-ups to name a few.

By "all other ops" I mean the operations which are not issued by NFS/CIFS/iSCSI/FCP (rest of the operations on a volume internally treated internal ops or in other words as other_ops.

Further classification of these "other_ops" is not done yet, this may be classified in future in at least two broader classification as internal ops and external ops. 

Thanks,