Choose hosting for web access, background work and files
Compare hosting options with the same application: portal access, a long-running PDF job and persistent files with controlled access need suitable services.
Choose hosting around the tasks your application performs. A platform that serves web pages does not automatically cover background work and persistent file storage. Record these requirements separately before comparing offers.
Record the three tasks of the same application
The following example is fictional. Customer portal P-46 displays project information after sign-in. PDF job J-46 should continue after the browser window closes. The resulting file F-46 must remain accessible later to authorised project members.
A separate database is planned for both options. Its connection, backup and recovery belong in the overall assessment. We compare a managed HTTP container platform with job execution and file services not yet confirmed, and a self-managed virtual machine with a disk. These are two example approaches, not complete provider offers.
Apply the same requirements to both operating approaches
The table records what must be clarified or provided for the example. It contains no results from a tested installation.
Component | Specific need | Managed platform | Self-managed VM |
|---|---|---|---|
Portal access | Signed-in people see their authorised projects. | Check the runtime, database connection and application permissions. | Operate the web server, runtime and database connection; manage application permissions. |
PDF job | Work continues independently of an open browser. | Plan a suitable job service or supported worker (background process). | Start and monitor your own worker and provide for safe retries. |
File storage | The file stays accessible after the application instance is replaced. | Connect persistent file storage with suitable access control. | Handle the disk, data backup and recovery yourself. |
A virtual machine can run these components together. That does not establish availability if the machine fails. Missing platform services must not silently be treated as included in the comparison.
Read the provider’s actual rules
Google’s Cloud Run container contract describes, for example, ephemeral local container storage and CPU allocation that depends on the operating mode. This concerns that product and its configuration. It implies neither a ban on all background tasks nor a limitation shared by every managed platform. Execution type, termination behaviour and persistent storage must actually fit your design.
Microsoft sets out the remaining cloud responsibilities for data, identities and access. Managed infrastructure still needs someone responsible for application code, permissions and the complete recovery path. Establish whether each task belongs to the provider, your own team or a service partner.
Check the whole workflow before choosing
In a suitable test environment with synthetic data, have a job started and close the browser. Interrupt and restart the worker. Check that the same job continues traceably and creates no unwanted additional result file. The application needs suitable job and retry logic for that; hosting alone does not provide the evidence.
Also replace the application instance and try opening the stored file with allowed and denied access. The platform becomes an option when the required services and connections are clarified. A self-managed VM becomes an option when the responsible team can provide operations and recovery. Then compare effort and costs for that complete scope.
Sources
Read more
All articles
Cloud & Operations with ASES
Hosting, deployment, maintenance and support