Putting all that (HEP-) Data to work - a REAL Implementation of an Unlimited Computing and Storage Architecture
Michael D. Ernst · 1996
Since Computing in HEP left the ‘Mainframe-Path’, many institutions demonstrated a successful migration to workstation-based computing, especially for application requiring a high CPU-to-I/O ratio. However, the difficulties and the complexity starts beyond just providing CPU-Cycles. Critical applications, requiring either sequential access to large amounts of data or to many small sets out of a multi 10-Terabyte Data Repository need technical approaches we haven’t had so far. Though we felt that we were hardly able to follow technology evolving in the various fields, we recently had to realize that even politics overtook technical evolution- at least in the areas mentioned above. The USA is making peace with Russia. DEC is talking to IBM, SGI communicating with HP. All these things became true, and though, unfortunately, the Cold War lasted 50 years, and-in a relative sense- we were afraid that ’50 years ’ seemed to be how long any self respecting high performance computer (or a set of workstations) had to wait for data from its ‘Server’, fortunately, we are now facing a similar progress on friendliness, harmony and balance in the former problematic (computing-) areas. Buzzwords, mentioned many thousand times in talks describing today’s and future requirements, including Functionality, Reliability, Scalability, Modularity and Portability are not just phrases, wishes and dreams any longer. At DESY, we are in the process of demonstrating an architecture that is taking those five issues equally into consideration, including Heterogeneous Computing Platforms with ultimate file system approaches, Heterogeneous Mass Storage Devices and an Open Distributed Hierarchical Mass Storage Management System. This contribution will provide an overview on how far we got and what the next steps will be. 1