Hidden VNXe performance statistics

Henri, my curious technical friend. The database of performance metrics is intentional - VNXe Engineering utilizes these statistics. There are many of us here who would love to get this information populated into Unisphere.The goal has to be balanced with priority of other features.

Nothing can be coded, tested and maintained without taking away from all the other areas. Voice your preferences, but realize there will always be more we want to do than we have time to do it in. I trust in my managerial colleagues to make the most appropriate choices as they prioritize asks.

My only additional note is to release that the team has not noted its existence in part because it can and likely will change over time. Keep that in mind as you decide how to utilize your R&D time. But while the information is available by all means use it.

All the best,

Matt

@mjbrender

Hey Giograves - could you send me detailed information about what you see in NetApp that you'd like to see in VNXe? I love to route those kinds of requests to the people that can make it happen. You can find me at [dot] at emc.

Best,
Matt

EMC seriously needs to get where the NetApp is in terms of perf counters from the command line.  It doesn't necessarily need to be in Unisphere. I do appreciate there is is at least something available, but very kludgie to get to it. The VNXe is surly to hit a low price point, but the right tools need to be baked in since these boxes can be slammed pretty easily. Lets hope

Hi Matt, thanks for your response.

I think first and foremost, customers should be able to "mine" perf data without need to "break" into encrypted (nar) or

sql files. Simple csv capable out files will do. I understand alot of this has to do with support revenue, but other vendors are leaving EMC behind in this area, with free powerful perf tools on mid-range frames.

Second, real-time counters that can give you a snap shot view of how busy your system is one of the truly powerful things with the ONTAP command line. With a simple command like "sysstat" you get a great birds eye view in real time of what's going on in the system, with self-explanatory column headers. Here are some examples.

CPU   NFS  CIFS  HTTP   Total    Net kB/s   Disk kB/s  Cache  CP Disk  FCP iSCSI   FCP  kB/s iSCSI  kB/s

And as we all know, when disk utilization hits above 60-70% latency starts to tank, and with a simple command like "sysstat -m", we can see potential problems quickly.

Netapp also has very good trace capability with the "statit" command. I learned the VNXe has something similar in the "svc_perfcheck" command, but the output doesn't roll up the averages for you and so if I wanted to run dozens of iterations over a long period of time like a day, I'd be generating way too much data that I can only use a graph to interpret. Also the svc_perfcheck's columns are not as intuitive.

Lastly, since the perf counters are some sort of "open" format, I can use Windows perfmon to connect and view stats.

Too much to ask for in a entry level box? Maybe? But if the goal is "unified", all that should really matter between VNX boxes is storage capabilities, not monitoring stats.

I would suggest looking at netapp's "sysstat" and "statit" commands.

For things like trending and correlation analysis, its perfectly understandable to have to purchase additional product like unisphere analyzer, or in Netapps case, DFM.

how about snmp ? for this product