OCUM, WFA and OPM Global Deployment

[david_ovenden]

Hi,

We are looking to monitor and manage our global estate.  We have 4 primary hubs, London, New York, Sydney and Hong Kong.  The administrators are distributed around the regions too.

I can see 3 possible options, all with advantages and disadvantages;

1. Central deployment - A single OCUM, OPM and WFA server.  This gives a single location to go to administer, but could potentially have slow access from remote hubs.
2. Dispersed deployment - an OCUM, OPM and WFA server in all 4 hubs.  This give better local performance but means 4 different connections to manage the estate
3. Hybrid solution - Single OCUM and OPM server, and 4 WFA servers.  This will give good local performance for WFA but not for OCUM...

Are there any best practices or recommendations around a global deployment, or does anyone have any opinions on which way we should be looking at?

[trentino123]

Hi David,

I would start with 1 WFA and many OCUMs ( 1 OCUM per region ).

The applications will have root/admin access to the controllers , when you have many WFA/OCUM, it adds complexity/workflow changes updates /access management, unless the WFA in the remote hubs only have root/admin access to their site only, which could be more secure.

Just my opinion to generate more thoughts.

[Adai]

Hi David,

              Please find my response with the procs and cons of each deployment model. Based on what suits your environment pic the one.

---

1. Central deployment - A single OCUM, OPM and WFA server.  This gives a single location to go to administer, but could potentially have slow access from remote hubs.

PROS:

=====

Most preferred way, as it gives one single configuration to be done on each product.

Consolidated view of your entire NetApp landscape at any given point of time.

No Extra or Additional load on the controller. Due to redundant monitoring.

Will give you the ability to view protection informations which span across geographies as they are all monitored by same server.

CONS:

=====

If your remote networks are unreliable you will get false alerts as complete monitoring is based on IP.

@Perceived slow response, but could be over come by accessing a remote Desktop @ the site where the management server is running.

Conclusion:

========

The only reason for not selecting this deployment model is if you hit CONS # 1.

---

2. Dispersed deployment - an OCUM, OPM and WFA server in all 4 hubs.  This give better local performance but means 4 different connections to manage the estate

---

@david_ovenden wrote:

Hi,

We are looking to monitor and manage our global estate.  We have 4 primary hubs, London, New York, Sydney and Hong Kong.  The administrators are distributed around the regions too.

I can see 3 possible options, all with advantages and disadvantages;

1. Central deployment - A single OCUM, OPM and WFA server.  This gives a single location to go to administer, but could potentially have slow access from remote hubs.
2. Dispersed deployment - an OCUM, OPM and WFA server in all 4 hubs.  This give better local performance but means 4 different connections to manage the estate
3. Hybrid solution - Single OCUM and OPM server, and 4 WFA servers.  This will give good local performance for WFA but not for OCUM...

 PROS:

=====
Local Access would be faster, but not faster than what the server can perform.

CONS:

=====
No single consolidated view
Need to write or use other tools to consolidate and report
Need to configure the same set of alerts, annotation in every management server
Need lot of VM's, associated power, cooling, space and increased cost.

Conclusion:

========
I would never go for it unless the global monitoirng wont work due to unreliable network, causing false positives

---

3. Hybrid solution - Single OCUM and OPM server, and 4 WFA servers.  This will give good local performance for WFA but not for OCUM...

 PROS:

=====

Dont See much

CONS:

====

Replicating or manging the contents in each WFA server. 

Always having a central automation tool will help a lot in dealing with space related contentions or thresholds.

The workflows in each WFA server will need to be restricted to a set of cDOT clusters to avoid, taking actions on the other site one.

This will be a developer night mare to keep updating the same workflows with four different finders etc.

Conclusion:

=======

I dont see any merit in doing this.

Hope this helps, BTW this is my personal opinion, based on many of this similar activity that I have done for NetApp customer in the past due to scaling issues with DFM, protection manager, provisioning manager and performance advisor.

Regards

adai

[bobshouseofcards]

One more variation to consider:

1 OCUM, 1 WFA, multiple OPMs (1 per hub).

Reasoning - central administration through single OCUM/WFA.  OPM queries, especially if there are many objects, can present a lot more traffic so collecting locally at the hub just works better, especially in wide ranging collection points.  OCUM is deisnged to rollup many OPMs so you have the central monitoring point with drill down to the individual performance details as may be needed.

I've alos noticed personally that OPM tends to run out of gas faster than does OCUM as the number of Filers/Clusters/Objects being monitored increases, probably because it seems OPM pulls and analyzes a lot of data per object.

Have two geographically dispersed locations in the US and one UK location - going with a local OPM at each site and central OCUM/WFA myself as next step from centralized everything, as OPM is the only current pinch point in our monitoring/administration.

Bob Greenwald

[Adai]

Hi Bob,

           1 UM, 1 WFA and 4 OPM (1 per site ) is really a good opiton. Yes OPM runs out of juice as it does have to do with collection, and number crunching for analytics.

But from what I heard the scale of OPM 1.1 is much better than 1.0 and they are constantly increasing the scale over releases.

The point that one OCUM can roll up multiple OPM negates some of the CONS of distributed models.

Regards

adai