Welcome to Community Week 2025! Join us to learn, connect, and be recognized as we celebrate the spirit of Community and the power of AI. Get the details  

Datacenter selection for midserver installation

Balakrishna_ABK
Tera Guru

We are implementing discovery, as per discussion with customer we got around 6 on prem data centers (1 main datacenter) and 2 azure cloud datacenters in different locations.

1.For on prem how to choose which datacenter to install midserver?

2.Is it ok to install 1 midserver for all 6 on prem data centers?

1 ACCEPTED SOLUTION

Pratiksha
Mega Sage
Mega Sage

For non prod you can go with 1 in each data center. For prod try to have two at least to form the cluster. This is with the assumption that mid server will be able to carry load for the devices. 

 

You should check how many devices you have. 

 

  • A MID server can process about a probe per minute per thread, and there are about 7 probes per device on average.
  • Both of those may vary, depending on the mix of devices (for example, it takes more probes to discover Windows servers than it does for routers) and the speed of your network communications and MID Server.
  • If the IP ranges are unusually large, i.e., 100 or bigger and mostly empty, the MID server will spend a significant amount of time scanning non-responding IP addresses, slowing the performance.
  • Suppose you have a location with about 2,500 devices (a mix of servers, network gear, printers, and workstations) all together. Further suppose that you wanted Discovery to do its job between 6 pm and midnight on Friday and 3 am on Saturday each week – a “Discovery window” of 6 hours.
  •  
  • Below is a rough estimate to see if a single MID server configured to use 25 threads could achieve this goal.
  •  
  • Using the rules-of-thumb, discovering 2,500 devices would require about

    7 * 2,500 = 17,500 probes

    With 25 threads, the MID server can process about 25 probes per minute.

Check mid server best practices on Nowcreate. 

 

Hope it helps. 

 

Regards, 

Pratiksha

View solution in original post

2 REPLIES 2

David104
Tera Guru

The answer to your question is dependant on a few factors.

 

As a general rule, you want the MIDs to be close to the workload that they are managing, this is because historically, introducing too much traffic between data centres can cause unwanted congestion on those links and latency might have an impact on performance of your discovery process. Most of the time, that doesn't seem to be as much of a concern these days. So, using one MID server (or preferably MID Cluster) for all of your on-prem data centres could be OK if hte network infrastructure supports it.

 

The second main consideration is whether these data centres are firewall separated. If you are trying to discover behind a firewall, then there are a whole lot of ports that need to be opened up for the MID server. This is not usually desireable as you are weakinging the security, adding management overhead for those rules and potentially contributing to congestion on the firewall. Planning MID servers to have a presence in all of you network segments is probably a better approach to avoid this.

 

Something else to consider though, what if you have an outage on the datacenter where your MID is. CMDB and discovery can have a part to play in a DR recovery scenario, so consider having MID servers at least on more than one of your data centers

 

Cheers,

David

Pratiksha
Mega Sage
Mega Sage

For non prod you can go with 1 in each data center. For prod try to have two at least to form the cluster. This is with the assumption that mid server will be able to carry load for the devices. 

 

You should check how many devices you have. 

 

  • A MID server can process about a probe per minute per thread, and there are about 7 probes per device on average.
  • Both of those may vary, depending on the mix of devices (for example, it takes more probes to discover Windows servers than it does for routers) and the speed of your network communications and MID Server.
  • If the IP ranges are unusually large, i.e., 100 or bigger and mostly empty, the MID server will spend a significant amount of time scanning non-responding IP addresses, slowing the performance.
  • Suppose you have a location with about 2,500 devices (a mix of servers, network gear, printers, and workstations) all together. Further suppose that you wanted Discovery to do its job between 6 pm and midnight on Friday and 3 am on Saturday each week – a “Discovery window” of 6 hours.
  •  
  • Below is a rough estimate to see if a single MID server configured to use 25 threads could achieve this goal.
  •  
  • Using the rules-of-thumb, discovering 2,500 devices would require about

    7 * 2,500 = 17,500 probes

    With 25 threads, the MID server can process about 25 probes per minute.

Check mid server best practices on Nowcreate. 

 

Hope it helps. 

 

Regards, 

Pratiksha