Databases Reference
In-Depth Information
16.3
Replicated Data Allocation—“All-beneficial Sites”
Method
The “all-beneficial sites” method selects all sites for a table allocation where the benefit is
greater than the cost for one additional copy of that table. You are assumed to start with
no copy or one copy of each table.
The benefit for an additional copy of a given table tab0 at site S is measured by
the difference in elapsed time between a remote query (i.e., no replicated copy) and a
local query (i.e., replicated copy available), multiplied by the frequency of queries to
table tab0 originating from site S.
The cost for an additional copy of a given table tab0 at site S is the total elapsed
time for all the local updates for table tab0 from transactions originating at site S, plus
the total elapsed time for all the remote updates of table tab0 at site S from transactions
originating at other sites.
16.3.1 Example
Let us illustrate the application of this method with a simple example of a database
schema and its processing characteristics. The average disk input/ouput (I/O) times are
given for a query or update originating from the same site in the network (local) or
combined disk and network service times from different sites (remote).
System Parameters
Table
Size
Average
Local Query
(Update)
*
Average
Remote Query
(Update)
*
tab1
300 MB
100 (150)
500 (600)
tab2
500 MB
150 (200)
650 (700)
tab3
1.0 GB
200 (250)
1,000 (1,100)
* Time in milliseconds.
User transactions are described in terms of their frequency of occurrence, which
tables they access, and whether the accesses are reads or writes.
