You will need to provision a dedicated Linux server or Linux Virtual Machine (VM) to run the API endpoint on. We have multiple customers who run a Windows-only server set up, and they have still been able to install our software by running a Linux VM on a Windows server.
It is not recommended to use an existing server that is already running other applications.
The total number of servers you will want available will depend on a few factors:
- The total number of environments you want to install in (e.g., UAT, Production, etc.)
- This changes from customer to customer. Customers typically get 2 servers when they believe that their testing processes might impact response times for production users. These organizations are testing heavily at the same time as many production users are interacting with dashboards.
- Customers typically get 1 server when they believe that testing will not impact end users. They usually do their testing outside business hours or do not have very intensive QA processes.
- How many servers you want for scalability/disaster recovery
- Our customers often allocate extra servers solely for disaster recovery. For more detailed information about our products' concurrency capabilities, please reach out to your Narrative Science representative and ask for the Benchmarking & Scalability Guide.
Using the recommended four-core deployment, the Quill Extensions API can successfully handle up to 100 concurrent requests.
Taking into account an average client’s usage of the installed Quill Extensions API, we find that clients often don’t need to scale vertically (i.e., by provisioning larger servers with more CPUs/memory) or horizontally (i.e., by adding more servers and placing them behind a load balancer) beyond our recommended system configuration.
However, if necessary, the Quill API can be installed in a multi-node deployment. Provisioning and configuration of the load balancers is the responsibility of the client. Please see Technical Architecture to reference diagrams for single and multi-node installations.
For more information on concurrency and scalability, please see the Narrative Science Benchmarking and Scalability Guide and reach out to Narrative Science if you have further questions. If you do not have access to this Guide, your Narrative Science representative can pass it along.
Please proceed to the next step: 2.0: Hardware & OS Specifications