Are there general sizing guidelines and best practices that can be use for setting up Process Discovery and managing performance?
Guidelines for assigning PD Agents to PD groups:
From different business units, PD Agents should be assigned to separate PD groups. Typically, there is one PD group per a line of business. Having such separate PD groups significantly reduces their analysis duration and hardware requirements for PD Analyzer. Note the more data recorded into the PD groups in total, the more hardware resources PD Analyzer will need to process the data.
The analysis is performed by PD Analyzer for each PD group individually. Analytics dashboards as the result of the analysis display information for each selected PD group as well as for all PD groups.
Hardware recommendations for PD Analyzer:
Running PD Analyzer on a dedicated server is highly recommended as PD Analyzer tries to consume all CPU resources at specific phases of data analysis. When PD Analyzer consumes 100% of CPU, other processes on the machine might be starved of resources.
To effectively execute analysis of the data recorded by 300 PD Agents, PD Analyzer should run on a powerful server. The recommended hardware is at least:
- CPU Intel Core i7 12 Cores 3.8 GHz (or equivalent)
- 64 GB RAM
Guidelines for databases:
For database sizing, the agent records between 300 to 700MB of data per user/day.
Each PD group has its own DB. PD documentation says: "We recommend using a dedicated server with fast disk drives, such as SSD for your MySQL database and run no more than 100 Agents simultaneously for one Process Discovery group". If the hardware is not powerful enough, we recommend that less than 100 PD Agents are assigned to one PD group. Otherwise, the number of PD Agents per a PD group depends on intensity of user activities and applications such activities are performed in. To sum up, running database servers on dedicated machines is highly recommended (i.e., one DB server per a machine). Note that DB for PD Analyzer should also be hosted on a dedicated machine due to a large amount of data it usually stores.
So, in case of 300 PD Agents, they might be organized, for example, into 9 PD groups (i.e., one business unit per a PD group), in other words:
- PD group A (50 PD Agents) - DB server 1
- PD group B (50 PD Agents) – DB server 2
- PD group C (50 PD Agents) – DB server 3
- PD group D (50 PD Agents) – DB server 4
- PD group E (30 PD Agents) – DB server 5
- PD group F (20 PD Agents) – DB server 5
- PD group G (15 PD Agents) – DB server 6
- PD group H (15 PD Agents) – DB server 6
- PD group I (20 PD Agents) – DB server 6
PD Analyzer will have its own DB server besides:
- PD Analyzer – DB server 7
Hardware recommendations for databases:
- CPU Intel Core i5 4 Cores (or equivalent)
- 16 GB RAM
- Fast disk drives (e.g. SSD)
- Fast network with large bandwidth
For performance monitoring:
- Review the Troubleshooting section located in the Online Help:
- Review database performance tools and look for expensive queries. Review with a database administrator for potential index enhancements where applicable.
Level of Complexity
Article # 3046317