Out-of-path, content-addressed writes with untrusted clients

A system for writing client data to a data server using a metadata server comprising a non-transitory computer readable storage medium and a processor executing instructions for:

• receiving at the metadata server a BEGIN_WRITE request from a client, wherein the BEGIN_WRITE request includes a plurality of hashes of data to be written to the data server;
• consulting a deduplication table to determine the data has not previously been written to the data server;
• transmitting both a list of data locations on the data server and a set of authentication tokens for the data locations to the client, wherein the authentication tokens authorize the client to write the data identified by the hashes to the data locations, wherein the authorization tokens are associated with an authorization timeout;
• receiving at the metadata server an END_WRITE request, wherein the END_WRITE request indicates the client has finished writing the data to the data server;
• subjecting the data written to the data server to a quarantine, wherein additional clients can not write to the data locations during the quarantine;
• verifying from the metadata server that the data is present on the data server when the authorization timeout has expired;
• and
• removing the quarantine in response to a passing verification.

A computer program product for writing client data to a data server using a metadata server, the computer program product comprising a non-transitory computer readable medium encoded with computer executable program, the code enabling:

• receiving at the metadata server a BEGIN_WRITE request from a client, wherein the BEGIN_WRITE request includes a plurality of hashes of data to be written to the data server;
• consulting a deduplication table to determine the data has not previously been written to the data server;
• transmitting both a list of data locations on the data server and a set of authentication tokens for the data locations to the client, wherein the authentication tokens authorize the client to write the data identified by the hashes to the data locations, wherein the authorization tokens are associated with an authorization timeout;
• receiving at the metadata server an END_WRITE request, wherein the END_WRITE request indicates the client has finished writing the data to the data server;
• subjecting the data written to the data server to a quarantine, wherein additional clients can not write to the data locations during the quarantine;
• verifying from the metadata server that the data is present on the data server when the authorization timeout has expired;
• and
• removing the quarantine in response to a passing verification.