cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
511
Views
0
Helpful
5
Replies

large scale ciscoworks 3.1 under vmware esx 3.5

craig.schnarrs
Level 1
Level 1

We are planning on upgrading our ciscoworks 2.2 installation. we are

going to lms 3.1 and are looking at running it under vmware ESX 3.5 sp2.

we would like to stay on a single server if possible. and we will not be utilizing HUM.

given that how many devices can we manage on a single box , and what caveats are there

when running under vmware ESX 3.5 sp2

also if the solution requires multiple servers. are multiple servers supported under

vmware.

we currently want to manage between 3000-4500 devices. from a technical point of view is

this feasable under vmware?

the devices are not all in one area they are dispersed geographically across the country,

please also point me to how i would size ciscoworks from a technical point of view.

thanks for any and all help!

5 Replies 5

Joe Clarke
Cisco Employee
Cisco Employee

You can certainly do this with VMWare 3.5 ESX and LMS 3.1. You just need to make sure your server meets the minimum requirements. For this device count without HUM, you would need a four-CPU machine. Each CPU would need to be 2.6 GHz or faster (or equivalent multi-core CPUs). You would need at least 9 GB of RAM (1 for the hypervisor, and 8 for LMS).

This configuration would require you to run Windows 2003 Enterprise Edition with PAE enabled.

This configuration would give you anywhere from 1000 to 2000 devices of growing room on a single server. That said, you will probably hear some recommendations about splitting this up to avoid the eggs/basket problem. The logical split may also help with network traffic when it comes to polling devices in geographically diverse regions.

do you know of any tech tips, or lessons learned regarding running lms 3.1 under vmware? I need the recovery ability provided by snapshots, ( because i like sleep and want to go home and cant till the problem is fixed)...

could i put more memory in? say 16 gb, how many devices should i run on a single server to be safe?

I don't think we've been supporting VMWare long enough to have a good list of leading practices. What seems to be the case with LMS in general, though, is memory and CPU make a huge difference when it comes to performance. As for LMS on Windows, it is generally advisable to do a few things prior to installation:

* Set the TEMP and TMP environment variables to some thing short (e.g. C:\WINDOWS\TEMP).

* Disable anti-virus software.

* Login on the console as a local Administrator.

* Install from locally mounted original media.

With Windows 2003 EE, LMS can support up to 32 GB of RAM. LMS 3.1 supports up to 5000 devices on a single VM. However, that gives you no room to grow. Your idea of 3000 to 4000 on a single server does give you good head room.

wow, thanks for all the advice!!

I can see how disabling the antivirus software would help, but isn't that dangerous?

and how do you think performance would be over a wan? the general idea is that the server would be at our main data center in maryland, and we are in warren michigan ( suburb of detroit) and we have about 1500 pieces of equipment in detroit. and the other 2500 are dispersed geographically across ths us ( i dont know details yet)

but what they are big on, is using user tracking.. they love it..

would you think lms 3.1 should be on its on server or share it with other vm's? ( im not super familar with ESX , but i dabble a bit)

as for performance , with 3500-4000 devices on one server and doing user tracking, do you think i could get away with one VM to do this?

thanks

You would only disable AV for installation. You can enable it once the product is fully installed. You should disable it every time you install patches, however.

While polling isn't all that intense (at least by default), things like software distribution do take a lot of bandwidth, and would benefit from being closer to the devices. The good news is that we do have some features in LMS to mitigate WAN bandwidth consumption. One is our remote syslog collectors. This will allow you to setup distributed syslog collection and filtering points which will eliminate syslog noise from the WAN. Another is software remote staging. Image deployment is probably the biggest single bandwidth killer in LMS. With remote staging, you can copy an image once to a host or server on your WAN close to a set of devices to be upgraded. Those devices can then fetch the image locally instead of going back across the WAN.

LMS requires a dedicated box. That is, the resources I spelled out for you need to be dedicated for the LMS VM. If you want to run other VMs, you will need additional resources. Only LMS and ACS VMs were tested on the same bare metal.

Yes, you can get away with one VM on one server with this device load provided you met all the other requirements.

Getting Started

Find answers to your questions by entering keywords or phrases in the Search bar above. New here? Use these resources to familiarize yourself with the community:

Innovations in Cisco Full Stack Observability - A new webinar from Cisco