You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
We use are own protocol called rtransfer - Oneproviders open multiple channels between each other and transfer the data in parallel (chunked). If configured properly, it is able to utilize most of the actual link throughput.
No, Oneclient typically reads or writes data in small blocks and requires another protocol with different optimizations, called ProxyIO. Due to the round trip delays on each block it is not viable for HPC, you should use DirectIO instead (when the Oneclient has direct access to the storage systems and contacts Oneprovider only to find the file location and metadata).
Currently, the data is proxied through the intermadiate Oneprovider. So, if you connect with a Oneclient to ProviderAlpha and ask for a file that is located in ProviderBeta, the request will be proxied and the data will be returned via ProviderAlpha. Our roadmap for the future includes the ability of Oneclient to switch betweeen Oneproviders in the background and get rid of proxy requests.
Closing the issue for now, let us know if you have any further questions.
I'm learning about onedata at the moment an I have a few questions about remote transfer performance.
The text was updated successfully, but these errors were encountered: