Permanent Standby connection to other servers

[StanleyK]

Hi,

is there any possiblity to connect one Permanent Standby server to more than one Main servers?

Thanks.

[AndrewScott · Accepted solution]

Yes, providing you buy the requisite number of licences. See the help for details:

https://tprojects.schneider-electric.com/GeoSCADAHelp/Geo%20SCADA%202020/Default.htm#DatabaseManager... 

[BevanWeiss · Accepted solution]

As mentioned, using the Database Manager you can have multiple instances of GeoSCADA Expert (or ClearSCADA) running on a single computer/server.

Each of these instances can be configured independently, so you can have each instance configured as either a Lone Server, or Hot-Standby Server, or Triple Standby Server, or Permanent Standby Server, or DMZ Permanent Standby Server.

Based on firewall rules and your particular network architecture not all of the combinations would likely make sense (i.e. having one instance configured as a Hot-Standby Server, and another on the same server as a DMZ Permanent Standby Server would be unusual... possible, but unusual).

You do however need the appropriate licencing for it however.  And it can end up a little tricky having easily rememberable port settings for them all, since they need to be unique (this applies for both the server port 5481[default] as well as the HTTP/HTTPS 80/443[default].. this default combo would work for instance #1, but instance #2 would need to have different numbers for these three entries)

[StanleyK]

Hi,

I know it can be a little bit tricky to maintain such a number of ports.

Unfortunately I need to create thirteen instances of Permanent Standby servers on a single machine (13 Permanent Standby server licenses).

So I have to handle 39 different ports. 

[BevanWeiss]

You might start to suffer some threading performance issues at 13 instances on the one server.

Unsure what your intent is exactly, but if you have 13 instances each with a number of threads, this is significant CPU contention, and you may find that you end up with significant overhead just from continual context switches.

If you have a server with lots of CPU cores, this would mitigate the issue, but you would likely have less challenges if you segmented the instances up and deployed across several physical servers.  Such CPU contention would only be amplified if this is also a virtualised server (in which case I would not recommend running other VMs on this physical host unless you have something like a Threadripper CPU).

Normally the reason for having permanent standby servers is to offload client connections from the main/standby servers.  It also has some cyber security benefits, since you can more tightly restrict connections to the main/standby servers.  With so many instances on the one physical server, your likely to have performance issues (hence taking away the benefit for the first reason), and you've created a critical failure point for numerous systems (taking away the benefit for the second reason).

Having a single server with a large point count licence (e.g. 13* 5k = 65k points in one licence) is cheaper than having 13 servers each with 5k point licences, but not all that much cheaper.  And it certainly has additional maintenance challenges.

I'd suggest trying to cluster instances if you can.  So rather than 1 server with 13 instances, or 13 servers each with 1 instance.  Perhaps have something like 4 servers, each with about 3 instances (and one with 4 instances). Or 3 servers, each with about 4 instances (and one with 5 instances).

This way:

1. Your N+1 hardware redundancy (cold spare server) cost is reduced.. since you don't need to buy a mega-spec spare server.
2. If you have a hardware failure, you don't lose all of your permanent standby instances, only a subset.  If you use virtualisation, you could even spin up the downed server images on your remaining physical hardware (particularly applicable if you had an even number of physical servers).
3. You don't have to come up with quite as many unique port numbers