Pega Server Architecture
We are using Customer Service, having 1500 active users, 110 agents, 32 queue processors, 94 job schedulers, calling 12 rest connectors , providing 4 rest service , 60+ active scheduled reports.
How we construct node classification?
Proximus
BE
Hello,
What's you PEGA version? I see tag above for 8.7.4 but I guess this is your expected version, not the current one since from PEGA 8, node classification is mandatory.
Such question would actually require much more info & details to get a clear answer.
How many servers have you got today? To which version are you planning to Go?
Node classification really depends on your design, expectations, globally your business as well as the targeted version.
Please provide some more info and this can high level be answered with some ideas.
Regards
Anthony
Turkish Airlines
TR
Thank you for your reply.
We are working on 8.2.7 and upgrading to 8.7.4
what kind of info should I provide?
moj.gov.sa
GB
It depends upon your VM specs = CPU and RAM as well.
if one node suppports 150-200 users - then you need 8-12 web nodes , 2 batch nodes , 2 stream nodes , 2 seach and reporting service
and if you want external separate nodes for hazelcast and Kafka - 1 node each may be required.
In Customer service are you using Customer decision hub as well , then you might more nodes as well..
Proximus
BE
You're in 8.2.7 without node classification?
from PEGA8, you need to mention a classification.
Or is it all universal? How is it structured today?
I checked documentation for 8.2.1 and it was already to be done, but mostly it is required for Stream Nodes.
The purpose of node classification is the give one or more role to your server.
This way you won't have everything started by default everywhere for which someone needs to stop when not needed.
As such, for servers where you want only users, you'll put webuser. Then your load balancer needs to be defined to redirect users to these webuser node only. With webuser classification only, there will be no background activities OOTB started.
Out of the box you need at least stream, backgroundprocessing & search.
- Stream for Queue Processor (there are some required OOTB ones)
- Search for your search nodes
- Backgroundprocessing for OOTB agents
Then for your own business, I would suggest not to use backgroundprocessing.
You can define your agents to run on Custom1, Custom2, etc. You can also create your own classification.
Last, your agents, QP, JS need to be assigned to the expected node classification.
You're in 8.2.7 without node classification?
from PEGA8, you need to mention a classification.
Or is it all universal? How is it structured today?
I checked documentation for 8.2.1 and it was already to be done, but mostly it is required for Stream Nodes.
The purpose of node classification is the give one or more role to your server.
This way you won't have everything started by default everywhere for which someone needs to stop when not needed.
As such, for servers where you want only users, you'll put webuser. Then your load balancer needs to be defined to redirect users to these webuser node only. With webuser classification only, there will be no background activities OOTB started.
Out of the box you need at least stream, backgroundprocessing & search.
- Stream for Queue Processor (there are some required OOTB ones)
- Search for your search nodes
- Backgroundprocessing for OOTB agents
Then for your own business, I would suggest not to use backgroundprocessing.
You can define your agents to run on Custom1, Custom2, etc. You can also create your own classification.
Last, your agents, QP, JS need to be assigned to the expected node classification.
When all this is done, based on your number of servers, the expected performance, etc, you "distribute" the classification (consider redundancy also, always better for safety purpose. Then also consider potential conflict of having same processes running on 2 servers also :-) eg: daily report at night which should run only once be defined on 2 servers :-) )
Then in some way, at the end, you'll have something like:
2 Stream (or more), 2 Search, 2 background, X for your own Custom node classification and Y servers for webusers.
A server can combine several roles.
I Hope you get a better view on what needs to be considered
Regards
Anthony