22 Introduction to the New Mainframe: Large-Scale Commercial Computing
2.4.1 Many servers
If data can be stored and processed on a single server, a large number of small
servers could be a more appropriate solution than a centralized data store. If
data needs to be accessed and updated from different servers, however, the
following problems have to be addressed:
Data retention may be required by law. When data is changed, does the
owning system or the updating system create a copy?
Because the data may need to be processed for both reading and updating, a
data locking mechanism has to be implemented and all servers processing
the data need to participate in this locking scheme.
Connectivity has to be established between the various ser