Video Screencast Help

Recommended server specs for NS 7.1 for a 10k client organisation

Created: 28 Sep 2013 | 6 comments

Hello all

We're struggling at the moment with a really poorly performing NS installation at my organisation.  At present we have around 2500 clients with the agent installed, and the NS server is struggling to say the least. Against advice, our server team provisioned this server on a VMWare virtual server. It has 4 virtual CPU cores, and it initially stated out with 16GB of shared memory allocated to it. I'm fairly confident this was underspecced for the number of clients we were going to place onto the system. We've since coaxed them into increasing the shared memory to 32GB, however this hasn't made a great deal of difference to the performance of the solution.

Now as a result of what I think is an underspecced sever, the console runs like a dog, making it painful to use. We now also need to roll out the agent globally to all of our other sites, bringing the client count to around the 10,000 mark.

My question is...  I realise our current NS server will not cope with 10,000 clients (it barely handles 2,500), and we'll likely want to move to a physical box with dedicated memory. What sort of specification of server would you recommend for an estate of 10,000 clients?

In addition to the NS server, each of our sites has a site server installed, including the headquarters where the NS box is located. These site servers run PXE services as well as package and task services. One issue we have is at our HQ, it seems that half of the 1,500 clients on this site pick up the NS as their task server, and the other half pick up the site server.  How do I force my clients to utilise the site server only? I'm assuming this will reduce traffic and load on the NS?  

Last question, for a site server running PXE, Package and Task on a site with 1,500 clients, what spec of machine would you recommend?

Thanks for reading and I look forward to your responses.


Ryan Spooner

Operating Systems:

Comments 6 CommentsJump to latest comment

RyanSpooner's picture

Sorry one more question...  we have an issue with tickle packets not reaching our clients. Could this be down to the performance of the NS or Site Server? This used to work fine, but recently when scheduling a job, the clients don't pick this up straight away and instead only pick it up on their next hourly check in. This causes havok with multi-task jobs as each task runs and then we wait an hour for the next one. Any thoughts?

Palvaran's picture

Hey Ryan,

A couple of questions, what is your SQL configuration?  Is it on box or off box?  If you it's on box AND virtualized that is the bulk of your problem right there.

Also, what about your site servers?  Are you just using the single NS to handle all tasks, packages, etc for all of the clients without any site servers to offload?

This guide was our biggest use when we rebuilt our structure too.

Starting from page 58 it gets into more nitty gritty of how you should layout and benchmark your system.

Systems Administrator
Rice University

Remember, "The happiness of your life, depends on the quality of your thoughts."

RyanSpooner's picture

Hi Palvaran, thanks for your reply.

We did have the foresight to put our SQL on an off-box physical server, although it is shared. That said, we've had our DBAs monitor the server, and it;s by no means under load, so I'm fairly sure it's not a SQL issue.

As mentioned in my opening post (which I forgive you for missing, it was a bit long winded), each of our sites has a separate site server on a 1U physical box (all sites apart from the HQ). The services are installed onto the print server box at each site, with some NetApp SAN disk presented to them for the package storage.

At our headquarters, the site server is hosted on a VM. This VM also has Workflow installed. This VM isn't anywhere near as powerful as the main NS VM.

Any thoughts please? I'll give your link a read now.

Palvaran's picture

There are just too many variables to figure out at the moment.  Your setup sounds like it should work, but you need more data to formulate a real analysis.  I know, it sucks and I hate the idea as much as you probably, but it might be time to call support.

The few flags that are going off in my head are the shared aspects.  What is being shared and what resources are they using.  In our environment, we were using shared storage with our SAN as well and noticed that it was slowing down our environment some.  In the link I sent, it mentions sharing and the specific IO needs that the disks that are being used need.  We had to rearchitect our striping and disk layouts to accomodate that.  Also, our VM guests had to be migrated off of some hosts in order to have more resources.  On the server side, we ended up tweaking IIS, recycling rates, and a ton of other stuff.

Here is one of the best links we are currently using.  We went to Vision this year in Las Vegas and Symantec gave a great presentation on what to do to improve your infrastructure.

In the end, after viewing that presentation, we went back and started to reengineer how everything was done.

One idea which we are going to pursue, on October 9th, the new version 7.5 will be released.  You could spin up a new VM and migrate your clients over.

Systems Administrator
Rice University

Remember, "The happiness of your life, depends on the quality of your thoughts."

HighTower's picture

What version of the platform are you running?  The most current is 7.1 SP2 MP1 rollup 7 (released last week).

Also, check out my previous post for a couple of tuning tips:

The biggest performance gains to be had were accomplished by fix our SQL config as well as getting really aggressive with daily index maintenance tasks in SQL.

RyanSpooner's picture

Thanks guys. I'm back in the office on the 7th (holiday this week) so I'll go through your suggestions then.

I'll also find out what specific version of the platform we're running. I know it's 7.1, but no sure of the patch level.

Thanks again.