We installed the RC for 4.0 yesterday, thought I would give a few initial impressions about the environment and process.
Our current environment is a main poller, 5 additional pollers and 1 additional web server, with a dedicated SQL server. We run NPM, NCM, NTA, UTA and VNQM. For the NTA 4.0 RC installation, we spun up a new VM for the NetFlow storage, 4 quad core processors, 32GB RAM and 1TB of storage.
For the upgrade, SolarWinds provided fantastic support. Jacob was online, as were several developers and support personnel. The install program (single image for all server types) was copied to all servers- I was initially unaware that it would be necessary to install on the storage server, but that presented no difficulties- just be aware, since that's where the install really begins.
Install there was a snap- except for our fire-walled environment. It was necessary to open tcp 1433 between the new data storage and the existing SQL server, as well as tcp 17777 from each of the pollers to the new storage server. We were not expecting that, so there was a brief delay while we prepared our environment. I don't see this as being a problem for production installs, as it will be documented, but it wasn't something I anticipated.
Once that was complete, we moved on the main poller, where again there were no significant issues except a licensing glitch (I'm temporarily now running an eval license) , nor on the additional web server.
The primary issues encountered were on my additional pollers. They must be running NPM 10.6. Although I recently upgraded to 10.6, we could find no evidence that I had and had to re-install that before we could do the NTA install. That was probably the biggest glitch we ran into, and didn't really have anything to do with the NTA installation at all. I must have just dreamed I installed 10.6 on all those pollers.:-)
Post install, my existing NetFlow data was migrated from the SQL server to the new data store. This is a process that took less than 5 hours, ran in the background with a neat progress banner telling me were it stood.
Performance. We monitor a bit over 2200 interfaces. Overall performance (total Orion system, not just NTA) is significantly improved. I guess not having all the flow data chucked at the SQL DB really helps. Reports I dared not run previously complete quickly. NTA performance is also significantly improved. Previously, we limited the default time frame to 1 hour to give decent performance when looking at NetFlow data. Now, a 24 hour view takes less time than a 4 hour view previously took to load. I am very pleased with the performance thus far, but I am interested in seeing performance once I have a full month of data in the database.
If you are in the RC, from what I experienced, I think you will be pleased with the results, assuming you have the resources to dedicate to a separate DB server. I'm still testing, looking for issues, but I have not identified any not listed above thus far.