Additional reference architecture needs
## Overview **DRI** - @tpazitny This epic is part of the work for the [Self-managed Scalability Working Group](https://about.gitlab.com/company/team/structure/working-groups/self-managed-scalability/) ## Description We would like to have additional reference architectures for sets of customers both larger and smaller than the 10k user size we based our first reference architecture off of. Based on [John's comment about ideal configurations](https://gitlab.com/gitlab-org/quality/performance/issues/15#note_189977500), we should consider starting with 2.5k and 25k architectures and iterating from there. This epic will cover a list of additional needed reference architectures and the steps to unblock creating them. ## Success * We have two additional reference architectures defined, tested, and documented. * One of these should support a smaller number of users than the existing 10k user reference architecture in order to be representative of a large number of our average self-managed customers. We are currently tentatively shooting for 2.5k users. * One of these should support a larger number of users than the existing 10k user reference architecture in order to be representative of a few large customers who are now or are considering using GitLab. We are currently tentatively shooting for 25k users. ## Tasks * Define a 2.5k reference architecture configuration * Define a 25k reference architecture configuration * Define success thresholds for our performance tests. * We are somewhat blocked from moving this epic forward by https://gitlab.com/gitlab-org/quality/performance/issues/7. To complete that issue, we will need to set thresholds for success for performance tests against the 10k environment. Those thresholds will be used within this epic while adjusting the environment configurations for the additional architectures. ## Questions - Are these user sizes appropriate representations for our intended purposes? - Do we need to determine personas for the types of users? e.g. 10k API-based users result in different loads than 10k UI-based users. - Do we need anything to be able to set up and maintain these architectures independently? (cc @grantyoung )
epic