Hi Sims, Welcome to Thwack
So in a Production environment, you will require a separate server for your Orion SQL Database. If you are also installing NTA then it has a requirement for an individual database, however, this database can be hosted on the same SQL Server and within the same instance as your existing Orion SQL Database. SolarWinds recommend that if you are sending a high volume of FPS (Flows Per Second) then you should consider moving the NTA Database to its own dedicated server.
Hope this helps
What is considered a high volume of Flows Per Second?
We run the Orion & NTA DB on the same physical server (which is quite beefy : 2x CPUs w/ 10 cores each and 768GB RAM and about 2.5 TB of storage) but we sometimes experience some NTA slowness. I wonder if this could be solved by moving NTA DB on its own server/DB instance since the hardware resources are definitely not an issue.
This is our current retention period and flow per second stats:
Out of interest how many polling engines do you have and how many elements are you polling? 768GB RAM for the SQL server is one of the largest I have come across. I've had 9 polling engines running off of 128GB and then 256GB RAM when it slowed slightly and it was fine, I considered that set up as Large at the time.
Don't forget that for SQL the biggest issue will be IOPS - So if you have that number of flows writing to the database then your causing congestion, which I suspect is why SolarWinds made it a separate DB!
That being said, I'm not sure what might help improve our performance:
1. Separate (smaller) DB servers
2. Instance stacking
3. Somehow increase IOPS (unlikely in the current setup, since we currently have SAN)
4. Performance tuning the existing server/DB
dgsmith80 is probably in a better position to say but your level of flows in high so a separate DB wouldn't hurt. However from a CPU and RAM spec your SQL box is incredibly capable.
How many web servers do you have? And as for disks are they SSD'd or 15k/10k? I'm guessing your raid is 10? and what Raid controllers?
This is a good base point for structuring Drives as well that will help...
Also simple things like..
Remove default web sites & application pools
Change directories for .NET temp and logs
Optimize compression settings
These are the official requirements..
As to specific raid controller it would all depend on what chassis you're running ect and how much through put you need. I'm defiantly not an expert in that area, so I would suggest creating a Thwack post about hardware and asking on there.
Oh completely! Its why they swapped back from fastbd (I think it was called?) to MS SQL again, great decision.
I'm just curious about the number for nodes and elements in general You might have seen higher spec'd SQL Servers but to me 768GB is a mega sized one. But then saying that I have only worked on a environments with 90,000 elements or less. Would be killer if it was Virtual as you often pay for RAM, like SQL is per sets or core's.
We currently have 5 polling engines (4 APEs), about 30K elements and the total DB size (Orion + NTA) is about 500GB.
The DB server is physical and we went with 768GB RAM (which might be a bit oversized at the moment) out of best practices considerations while also leaving some room for future growth (we expect to triple our managed node count - to a total of about 6K, which might more than triple the total element count - in the next 1-1/2 years).
Before NTA moved to SQL I think we were and about 320GB DB size and 512GB RAM and we had pretty decent performance. So, to your point, our performance issues might be tied to IOPS count constraints or other bottlenecks.
Based on this deployment guide SolarWinds NTA Flow Storage database deployment options I would suggest you need to separate. The SolarWinds Scalability guide says 50k FPS per polling engine and 300k per instance, but you might find more accurate information looking in the NTA Section of Thwack.
Thanks for the reply .
What is the hardware requirement for NTA server . Can I install it on the same NPM . ?
In the below link , there are some terms flow sources count ,nodes interface per node ,nodes ,endpoints count ,Flows per second per system
What are they , how it really matters
All depends on how many Flows you want to have sent to it. But yes you install NTA application on your NPM Server. You just need a separate DB on your SQL server. It can be the same instance as well with 2016 like David said. Just make sure your NPM server is 2016 or newer. I would assume it would be?
I think in smaller set ups Solarwinds are very generous on Server spec's. So personally I would keep with the same resource you already have on the Main polling engine (NPM Server) and then monitor it for performance. Strictly speaking its probs "text book" bad advise from me there but its real world advice IMO.
Solarwinds now include Orion Health checks as default so it will soon tell you if there are major issues. But If you're okay in resource now and not consuming high CPU, RAM and Bandwidth then you will be fine. Its more DB intensive anyway then Server, so having the right spec DB is slightly more important in my view.
I might be wrong on this as I never do it this way but I'm 90% sure the new installer actually creates the NTA DB, if you have full DB access when running the configuration wizard. So if its a smaller install its actually really easy to just wiz through it and not actually give too much though to resource. Please don't be that lackadaisical if you plan to hammer it with Netflow flows though.
SolarWinds solutions are rooted in our deep connection to our user base in the THWACK® online community. More than 150,000 members are here to solve problems, share technology and best practices, and directly contribute to our product development process. Learn more today by joining now.