EMC have released an OVF appliance that is meant to allow you to store and browse 30 days worth of VPLEX performance statistics. Version1 is limited to just a few metrics, but it is a very welcome addition to the VPLEX monitoring tools that are available! Requires GeoSynchrony 5.5

—-

Today I was looking up some information on vplex on the EMC support site, my eye was quickly drawn to the following entries:

I have seen no mention of this at all on either Twitter or on the VPLEX community space at EMC: https://community.emc.com/community/products/vplex

This is typical of EMC in my experience, they are terribad at disseminating support information and making new stuff ‘discoverable’.

So what is this thing?

Up till now, you had several ways to monitor, save, and analyze VPLEX statistics.

  • The GUI, but that only shows live data, no history, and only shows very few metics and only on high level
  • VPLEXCLI: Monitor create, Monitor collect, etc.  Powerfull CLI commands, any statistics can be saved. Can create exportable CSV files. But hard to use and understand, and for live monitoring the implementation is truly horrible, scrolling across your screen in a disruptive way, no ‘top’ kind of function here or anything
  • EMC VIPR SRM. EMCs statistics and analytics suite. Good for all kinds of EMC products, uses a  ‘perpetual’ version of the above mentioned monitor construct. But definitely not a free tool.
  • If you have VMware vROPS: EMC Storage Analytics.  Adapter for vROPS, but again not free. v3 of this adapter supports vROPS 6.x
  • SNMP. Vplex comes with a MIB, but my experience with it so far is that its got some serious compliance (and syntax) issues that are preventing it from working in, for example the vROPS SNMP adapter. (this was my attempt at a ‘cheapo’ EMC Storage Analytics’ 😉

So, nothing we had so far ‘just worked’ as a fast, and -free, gui-based way of seeing some deep statistics. There was something to be said for this not being available in the product itself. It looks like with “EMC VPLEX Performance Monitor” , which is a free OVF download, they are attempting to address this concern.

Lets check the release notes.

Product description

VPLEX Performance Monitor is a stand-alone, customer installable tool that allows you to collect virtual volume metrics from a VPLEX Local or VPLEX Metro system. It allows Storage Administrators to see up to 30 days of historical virtual volume performance data to troubleshoot performance issues and analyze performance trends.

The VPLEX Performance Monitor tool is delivered as an OVA (Open Virtualization Format Archive) file that you deploy as a VMware virtual appliance. The virtual appliance connects to one VPLEX system and collects performance metrics for all virtual volumes that are in storage views. Historical virtual volume metrics are stored in a database within the virtual appliance for 30 days. The virtual appliance has a web application which allows you to view the data in charts that show all 30 days of data at once, or allows you to zoom in on data down to the minute.

The VPLEX Performance Monitor charts the following key virtual volume metrics:

Throughput (total read and write IOPS)
Read Bandwidth (KB/s)
Write Bandwidth (KB/s)
Read Latency (usec)
Write Latency (usec)

Note: The VPLEX Performance Monitor can connect to one VPLEX Local or Metro system at a time. To monitor additional VPLEX systems, deploy a new instance of the tool for each VPLEX Local or Metro system you want to monitor.

Ok, so admittedly, for a version1, not all that much here, no port statistics or backend storage metrics for example. But in most cases, you are gonna be interested in your virtual volumes most of all anyway, so a good start.

Only 1 VPLEX system at a time? We have 2 Metro-Cluster setups in our environment, which translates to 4 engines in total. Does a ‘system’ equate to an engine? I think so, which means I would need 4 of these appliances. Oh well.

30 days is a nice sweet spot for metric saving as far as I am concerned. This appliance is using an embedded database, so don’t expect options to save your data for years. Get VIPR SRM if you want that.

IMPORTANT Version 1.0 cannot be upgraded. When the next release is available, you must delete the current VPLEX Monitor virtual appliance and deploy the new one. All performance data and user information will be lost.

  • The VPLEX Performance Monitor requires a VPLEX Local or Metro system running GeoSynchrony release 5.5 (VS2 hardware only).
  • The VPLEX Performance Monitor is not supported for use with VS1 hardware.
  • This version supports connection to a VPLEX system with a maximum of 4,000 virtual volumes.
  •  This release of the VPLEX Performance Monitor is not FIPS compliant. Contact EMC Customer Support if you encounter any issues installing or using the VPLEX Performance Monitor tool.

 

Take note of the GeoSynchrony 5.5 requirement. This version only came out recently, so I don’t expect many people to be running this yet.
We don’t in any case, so I can’t provide you with an install demo, yet :p

If you have GeoSynchrony 5.5, go give this a try:

https://download.emc.com/downloads/DL62040_VPLEX_Performance_Monitor_1.0.ova
https://support.emc.com/docu62030_VPLEX_Performance_Monitor_1.0_Release_Notes.pdf?language=en_US

(EMC Support account required)

Update 03:03pm: Was googling for “EMC VPLEX Performance Monitor” to see if anyone else had mentioned it yet, came across this video (with 20 views so far, wow!) that showcases the new tool.  https://www.youtube.com/watch?v=HiJgmbLkeTU

Tags: , , , , ,

One Comment to “EMC VPLEX Performance Monitor v1.0”

  1. Gary O says:

    Hi Robert,

    Excellent post! This VPM has been a much anticipated product release, and I totally agree with all the current challenges that come along with performance monitoring of VPLEX.

    I can clarify this paragraph:

    Only 1 VPLEX system at a time? We have 2 Metro-Cluster setups in our environment, which translates to 4 engines in total. Does a ‘system’ equate to an engine? I think so, which means I would need 4 of these appliances. Oh well.

    VPM will monitor one “system”. A system is defined as:

    – A Metro (so it will monitor two sides of a Metro simultaneously, showing you storage-view configuration and virtual-volume performance information from both clusters.) At each cluster: one, two, or four engines

    – A Local cluster – one, two, or four engines

    You definitely do not need a VPM to monitor each engine in a cluster, one VPM will take care of the cluster regardless of the engine count. (It aggregates statistics across the engines where appropriate – totals for IOPS and MB/s, and averages for latency.)

    So it’s one VPM per system. And so in your environment if you have two Metros, you will need two VPM instances.

    Gary

Leave a Reply

You can use these tags: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>