Server Sizing
We are new to Pyramid 2018 and are trying to get an idea on server sizing, in particular memory and cpu, before we lock in our configuration in AWS.
Our current config is:
Processor: Intel(R) Xeon(R) Platinum 8124M CPU @ 3.00GHz 3.00 GHz
Installed memory (RAM): 15.4 GB
System type: 64-bit Operating System, x64-based processor
For people who know something about AWS:
Instance Size: c5.2xlarge
Architecture: AMD64
I imagine we will have approximately 25 concurrent users at peak.
Does that seem a reasonable configuration?
Any help would be appreciated.
Thanks.
Kevin
4 replies
-
Hi Kevin,
Check out the Scaling Guide in Help, http://help.pyramidanalytics.com/Content/Root/Guides/Scaling/Scaling Pyramid 2018.htm
This document discusses in some detail the different Pyramid services and how they consume different resources. From this, you can start to determine a configuration suitable for your scenario.
Without knowing the number of CPUs, underlying database and other details, it is difficult to define exactly what your configuration should be.
Hope that helps.
Ian
-
Ian,
Thanks for pointing me to the scaling guide. We are closest to the Simple Standalone configuration. The server I described in my previous post holds all of the Pyramid 2018 servers including the in-memory database. We have a SQL Server 2017 server that is hosting the repository and most of our real data is coming from a SQL Server 2012 server. Since the Pyramid server is a Windows Server 2016 server in AWS I am not sure how to determine the number of cores. We are in the process of moving users over to Pyramid 2018 so we do not have much to judge the performance of our setup by.
Kevin
-
Hi Kevin,
Are you going to be making much use of the In Memory Engine? If so, we would recommend that had its own server.
Then what about high availability? If that is required, then at least two servers running all services would be needed.
I would perhaps consider more RAM, doubling it to 32 GB, and you will require 8 cores to install.
Every 3 months or so we run a benchmark test on the latest build to check that we are not losing performance.
From the last iteration, we get a figure of around 20 - 25 concurrent users per core when considering just the Runtime Engine. This is in a clustered Pyramid farm configuration, where the Runtime Engine has its own server.
Running everything on a single 8 core machine means of course that the cores are shared across all Pyramid Services, plus, of course, Windows itself.
I think it is reasonable to start with your current configuration, maybe increase the RAM a bit and ensure the core count is high enough.
Then monitor and adjust later if necessary. As it is a virtual machine, that should relatively easy.
Hope that helps.
Ian
-
Ian Macdonald The URL seems to be expired. This is the correct URL: http://help.pyramidanalytics.com/Content/Root/Guides/Scaling/Scaling%20Pyramid%202018.htm