NPM architecture for rather large expansion

Good morning,

I've been asked to get a quote for a 2nd poller this then drove another discussion with lifting shifting our existing Solarwinds platform to our datacenter and designing the platform 'properly' with the ability to provide access to external partners.

Currently we run the application server on a virtualised server (running on Nutanix AHV platform) and using a legacy DB server (SQL 2012).

As usual my budget is likely to be low (cost of 2nd poller and OS license to run it), another requirement discussed is to put the solution into a DMZ so remote access can be provided (my believe this is being driven by the desire to give our franchisee partners ability to see status of their sites that we manage).

We already have an HPE DL380G9 server in the DC that we're thinking of moving the solutions DB to, running 24GB of memory and direct attached storage of 300G on 15KRPM spinning disks and MS SQL 2K12 (i believe we have upgrade rights to 2014 but unlikely to 2016). note this is for the DB only and only if we move the solution to the DC.

License wise we bought the unlimited node license for NPM, SAM, NTA before SW changed its licensing back in 2010 or so ..

Currently we are monitoring

Polling rate 100


ELEMENTS	24233
NETWORK NODE ELEMENTS	5618
VOLUME ELEMENTS	9951
INTERFACE ELEMENTS	8664
UNDP POLLING RATE	0%
ROUTING POLLING RATE	5%
SAM APPLICATION POLLING RATE	25%
POLLING RATE	87%
SAM WINDOWS SCHEDULED TASKS POLLING RATE	0%
HARDWARE HEALTH POLLING RATE	25%
FIBRE CHANNEL POLLING RATE	0%
WIRELESS POLLING RATE	2%
VIM VMWARE POLLING RATE	0%
TOTAL JOB WEIGHT	4501
NUMBER OF HW HEALTH MONITORS	783
NUMBER OF HW HEALTH SENSORS	1904

I am aware that we are already over loading the existing solution

The driver is to add all tills and various other (generally windows) devices so we can do realtime auto remediation (this would be monitoring services, restarting services, anything else that could be done etc) and alerting to enable a complete proactive support package.

1/ Can I realistically get away with running the application servers / pollers on Nutanix AHV? given i am currently running the app server on Nutanix (and the sql server is ancient)

2/ Would using local agents on the devices being monitored lessen the load on the app/poller servers?

3/ How hardened is SW for use in a DMZ (I have some concerns over this idea)

4/ My assumption is this isn't a good / viable idea but can app and poller servers live on different networks (i.e. we have the 2nd poller in the DC with the app server and DB server remaining where it is).

5/ is there any specific integration into service-now ? Our support partner uses service-now so this might be a useful feature.

Is this getting to the scale I should realistically be taking to the SW Sales guys?

Regards

Nigel

Find more posts tagged with

architecture

npm

design

Accepted answers

All comments

brscott

1/ If you can run windows on it, then you can run solarwinds on it. Resources is going to be the key there.

2/ Yes, some, but not sure it's noticeable.

3/ GAK! Actaully I don't know, I just know I would not do it. At least get a separate additional web server just for external access. They are cheap.

4/ Actually that is a good idea, it will just need access to the DB.

5/ Yes there is integration to service-now. I tried using it, but unfortunately we don't use service-now properly, so I had to write my own integration. We apparently don't use most products the way they are intended... /sigh

The sales guys and gals LOVE to talk... and there are ALWAYS end-of-the-month/quarter/year special deals...

Yes, you do need another polling engine, then after that, you *really* need to upgrade the SQL, and make it a dedicated SQL server just for Orion. Then by next year you will probably be needing a 3rd polling engine. You almost need a 3rd one now, but your budget probably isn't going to support it, they are a bit pricey.

brscott

You actually have a nice balance of network/node/volume elements. So you can support a higher number of total elements than others of us. Those volume elements don't put as much of a load on the engine because of the long polling interval. Lucky you. Anyway keep an eye on the "polling completion rate", anything less than 100 (rounding up) means there was more work then it could get done, and there will be gaps in your graphs. Not sure how important that is to you.