@@ -210,8 +210,9 @@ Inactive users do not consume Coder resources.
210210
211211When determining scaling requirements, consider the following factors:
212212
213- - ` 1 vCPU x 2 GB memory x 250 users ` : A reasonable formula to determine resource
214- allocation based on the number of users and their expected usage patterns.
213+ - ` 1 vCPU x 2 GB memory ` for every 250 users: A reasonable formula to determine
214+ resource allocation based on the number of users and their expected usage
215+ patterns.
215216- API latency/response time: Monitor API latency and response times to ensure
216217 optimal performance under varying loads.
217218- Average number of HTTP requests: Track the average number of HTTP requests to
@@ -239,6 +240,14 @@ We recommend disabling the autoscaling for `coderd` nodes. Autoscaling can cause
239240interruptions for user connections, see[ Autoscaling] ( ../scale.md#autoscaling )
240241for more details.
241242
243+ ###Control plane: Workspace Proxies
244+
245+ When scaling[ workspace proxies] ( ../workspace-proxies.md ) , follow the same
246+ guidelines as for` coderd ` above:
247+
248+ - ` 1 vCPU x 2 GB memory ` for every 250 users.
249+ - Disable autoscaling.
250+
242251###Control plane: provisionerd
243252
244253Each external provisioner can run a single concurrent workspace build. For