It is perhaps easiest to consider ME: !,' to be a buffer for the mass

memory device. The memory system then can be considered to be a mass

memory system using relatively slow speed mass memory with a higher

speed core resident beffer. Access to stored data is made through the

buffer. The memory system handles all access to the mass memroy.

 Although to the user of the memory system it appears that all data

 are kept in MEMORY, an understanding of the interaction between MEMORY

 and mass memory helps in understanding the system. The random access

 data set has associated with it a logical record length (1,000 words in

 the case of the memory system). The data set is divided up into a

 number of these logical records. In this particular implementation,

 written for IBM 370/165, the random access data set has data written to

 or read from a particular record on a disk. This makes it possible to

 access a particular record. Since the records are numbered sequen-

 tially, the words in the first record can be thought of as being words

 1 to 1,000 and the words in the second record as 1,001 to 2,000, etc.

 Figure 5-6 illustrates this addressing scheme. If the address of a word

 of data is known, then the record of the data set it is stored in can

 easily be determined.

 In the memory system the first record is reserved for a special

function which will be described later and is never used for data

storage. Data storage begins with the second record. Data which are

moved to MEMORY have associated with them absolute addresses, which are

the word addresses that the data would have were they stored onto the

random access data set. This can most easily be visualized by imagining

that the vector MEMORY is superimposed on the random access file in Fig.

5-6. The first word of MEMORY would have absolute address 1,001, the