General Consideration: Atlassian Cloud is to be considered a moving target in the context of this document, which means features and capabilities can change daily, and so individual aspects of this discussion might be outdated at a later point in time.
There are 3 resources “in general” to Cloud vs. Data Center with Atlassian:
Cloud vs. Data Center Comparison by Atlassian – this page answers a lot of general questions. Some I will pick up and qualify further from our perspective:
Data Center Roadmap – Atlassian has now a dedicated Cloud Roadmap, which in a way, is also a longer-term commitment to offer customer an “on-premise” solution:
Many third-party apps only provide limited capabilities in JSM Cloud, as they are designed for Data Center. This is especially true for portal theming, where an integrated portal can be designed much more seamlessly using a variety of apps.
Integrations
Jira Cloud has better integration with SaaS tools, especially things like Atlassian's own OpsGenie.
Integrations to on-premises databases or systems are possible but often limited due to corporate security policies or technical challenges. We usually do not recommend integrations with on-premises instances unless it's strictly a pull-principle integration from the on-premises side.
Third-party system integrations can also be a blocker for cloud migrations.
Integrations with on-premises systems are the strong point of Data Center, as well as easier extendability using third-party apps or custom plugins.
Integrations to cloud systems often require complex firewall rules in order not to expose the internal instance to the outside world
User Access
Atlassian Cloud can be more accessible for large corporations using multiple networks or security zones, as it brings all users onto unified cloud infrastructure.
SSO features are more costly on a per-user basis due to the high pricing of Atlassian access. Due to the technical design, there is no third-party alternative for SSO.
SSO can be achieved with a variety of tools and methods and is generally cheaper.
Availability
Cloud Infrastructure is managed by Atlassian and is not reliant on "local" resources and their availability.
The availability of third-party apps needs to be considered here because they run on their own infrastructure. Overall system availability will not be the same compared to a DC instance, depending on how many third-party apps are in use.
Availability needs to be managed locally. A clustered environment provides benefits here in general for reliability, as well as the ability to do updates on individual nodes.
There is a risk of "rogue" third-party apps or external integrations bringing down the entire instance which is not present in the Cloud.
Federation
In Jira Cloud, different "Sites" can be created, where "Business" Resources could become owners in a closed, defined scope
While federation with multiple Instances is possible, there is obviously a bigger overhead in managing the instances and licensing. Also, the ownership per instance still requires very technical resources to maintain workflows and other configurations
Backup/Restore & Disaster Recovery
Vanilla Atlassian Products/Sites can be backed-up and restored. An automatic backup process to a company managed backup location can be automated with a script, but is not easy out-of-the-box
Backup and Restore of 3rd Party Apps have to be handled individually and are not implemented for most apps
Creation of a "Fall-back" on-prem instance is currently only limited possible and not an easy out-of-the-box feature
Backup and Restore of each Instance including all 3rd Party Apps is in full control of the company
Cloud is missing many useful out-of-the-box importers and integrations. There are items on the roadmap, but many are still missing.
What does work is the publishing of Insight Discovery into Cloud
Offers many out-of-the-box importers:
NVD (National Vulnerability Database)
AWS Cloud
Azure Cloud
Google Cloud
SCCM
Snow
Service Now
…
Scripting & Automation
Only functions via Jira Automation or REST API are available
Besides Jira Automation and other Apps that can interface with Insight, there is also the possibility to use groovy scripts to automate actions in Insight
Reporting
Reporting out of Insight is simply missing in the cloud. Atlassian has also made no committment to delivering the Insight App for Confluence
Insight comes with a basic reporting module that allows the creation of lists.
Also through the use of the Confluence Insight Macro, it is possible to create complex documentation like an Application Manual, e.g.: Single Confluence Page with several Insight Macros in it that show which Servers, Databases, Software and other assets are linked to the application.
OpsGenie & Jira Service Management Integration
The integration of Insight into OpsGenie and JSM is better in the Cloud. With exception to the OpsGenie Part, this can also be achieved on-prem
Insight Integration to OpsGenie is not out-of-the-box available on-premise.
Insight Recommendation
Call me “old-fashioned” but for larger organizations with a complex structure of assets, relations to other objects, etc., this information belongs on-premise. Not only for the richer features, but also for security and scalability. In general, if your Service Desk is on-premise, then the CMDB also belongs on-premise with the service desk.
For Glencore, putting complex cases as the Oracle Schema model for example, we will need most likely some scripting. In the cloud that will not be possible. Then having part of the processes in the cloud and others on-premise, with different CMDB-like schemas, will become a challenge.
→ Currently I recommend to stay on-premise with Insight for Glencore!
General recommendation
Choose Cloud for
Processes that can be designed and run with out-of-the-box product capabilities
Users that are new to Atlassian products and do not have requirements for features only available on Data Center
Hybrid collaboration situations with external companies
Small teams with very specific requirements in apps and customizations, that you do not want to have in your on-prem instance
Choose Data Center for
Processes that require connections to on-premises infrastructure
Workflows that connect to existing Jira DC processes and tools (avoid Jira to Jira sync situations)
