TeraScale SneakerNet: Using Inexpensive Disks for Backup, Archiving, and Data Exchange

Abstract

Large datasets are most economically trnsmitted via parcel post given the current economics of wide-area networking. This article describes how the Sloan Digital Sky Survey ships terabyte scale datasets both within the US and to Europe and Asia. We 3GT storage bricks (Ghz processor, GB ram, GbpsEthernet, TB disk) for about 2k$ each. These bricks act as database servers on the LAN. They are loaded at one site and read at the second site. The paper describes the bricks, their economics, and some software issues that they raise.