Best practice for MID Server Sizing?

[robertgeen]

Hello,

I know on the wiki there is an outline for what the minimum requirements are for sizing   MID server but are there any documents around sizing for discovery/service mapping? I'm assuming that an enterprise discovery would require more resources then that of a small business. Curious if anyone has anything they can share. Thanks.

[JBark]

Hi Robert,

I used the attached document when we initially rolled out SNow two years ago. I have 3 virtual servers hosting 3 Mid servers each with 3 load balancers setup with 3 MIDs each across the 3 hosts. We have 14000 CIs being discovered and it works just fine.

[Ankush13 · Accepted solution]

Excellent estimate was given by Doug in this post. We have to keep in mind that java will take memory and more & faster cores will always be a plus point.

[doug_schulze]

Here is my 'secret sheet' I use in my deployments..

[brukasmj]

Doug,

Myself and a colleque are testing Discovery in this infrastructure environment:

• Windows Server having the Mid server role
  • Windows 2012 R2
  • OS drive (C:) 40 GB
  • Data drive (E:) for the midserver application(s)

  • Processor(s):                           8 Processor(s) Installed.

    
    

  • Total Physical Memory:         16,383 MB

    
    
  • 5 midservers installed (4 for Johns instance and 1 for colleques' instance)
• PDI for John
  • 4 mid servers clustered for load balancing
  • configured OoB
  • Discovered 312 Computers, 30 switches/routers, 40 servers, 20 printers
• PDI colleque
  • 1 mid server
  • Configured OoB
  • Is discovering another building then John

I am using your great excel sheet for calculations and planning, based on the information described earlier it should be completed in 3/4 hours.

If we want to exluded the pc's / laptop for this discovery we only can do that by excluding the specific ranges for this right ?

We can't do it based on discovery as it uses wmi. My thinking is correct ?

We always will be adding complete ip-subnets assigned to a building. So this probably means the discovery takes longer then specifically adding individual ip addresses ?

I mean can't we use the behavior option to exclude ssh, snmp etc ? or is this not a smart thing to do ?

thanks

John

@Is it having impact using one dedicated server serving mid-servers for 2 different snow pdi instances ?