| Electronic discovery is a file system intensive process. Speed of processing can be critical. Discovery processes can take days to run to completion. A network failure may mean restarting the discovery process. The site used a SAN to which many Windows XP clients were directly connected. Processing was slow and prone to failed runs. |
The proprietary SAN was replaced with a number of replicated back-end storage servers. Each server has 64 TB storage. The storage is provided to front-end cluster nodes as iSCSI resources. Front-end storage nodes run Linux-based cluster servers. Storage is shared among multiple front-end systems using a clustered file system. Each front-end node runs Samba-CTDB to provide parallel CIFS (common Internet file services/MS Windows shares). Up to 200 Windows clients simultaneously connect to the cluster which appears to the Windows clients as a single large server. |
| The customer processes business data to produce dense-packed archives of medical test records, clinical trial information, and other sensitive information that often requires analysis and reporting. Processing time is important, but the critical requirement is that the back-end system is immune to Windows XP clients that suffer failed processing. |
The client chose to implement a collection of distributed storage servers and to provide a unified file system to Samba CTDB front-end nodes using GLusterFS. They required assistance with the design and implementation of the Samba CTDB cluster. The Samba CTDB cluster acts as an Active Directory domain member server. |
