Database Reference
In-Depth Information
Table 7.
Example for evolution paths
Storage Struc-
ture: 1st Evolu-
tion
Storage Struc-
ture: 2nd Evolu-
tion
Storage Struc-
ture: Initial
Mutation Rules
Mutation Rules
Event:
Sorted array=Full
Heredity based selection:
Workload=Read intensive
Data access=Unordered
Mutation:
=
>
Evolve (Sorted array
−
>
Sorted
list)
Event:
Sorted list=Full
Heredity based selection:
Workload=Read intensive
Data access=Ordered
Mutation:
=
>
Evolve (Sorted list
−
>
B+-Tree)
Sorted array
Sorted list of sorted
arrays
B+-Tree of sorted
lists(As leaf nodes
for data storage)
Event:
Sorted array=Full
Heredity based selection:
Workload=Read intensive
Data access=Ordered
Mutation:
=
>
Evolve (Sorted array
−
>
B+-Tree)
Event:
B+-Tree=Full
Heredity based selection:
Workload=Read intensive
Data access=Ordered
Mutation:
=
>
Evolve (B+-Tree
−
>
HLC
(B+-Tree based))
Sorted array
B+-Tree of sorted
arrays(As
HLC of B+-
Tree(As leaf nodes)
leaf
nodes
for
data
storage)
Event:
Sorted array=Full
Heredity based selection:
Workload=Write intensive
Data access=Unordered
Mutation:
=
>
Evolve (Sorted array
−
>
Heap
array)
Sorted array
Heap list based on
heap array muta-
tion rules
Event:
Heap array=Full
Heredity based selection:
Workload=Write intensive
Data access=Ordered
Mutation:
=
>
Evolve (Heap array
−
>
Heap list)
&
Generate (Secondary index = Sorted
list)
Event:
Heap list=Full
Heredity based selection:
Workload=Write intensive
Data access=Ordered
Mutation:
=
>
Evolve (Heap list
−
>
Hash
table) &
Evolve (Secondary index = Sorted
list
−
>
B+-Tree)
Heap array
Heap list
Hash table
Event:
Heap array=Full
Heredity based selection:
Workload=Write intensive
Data access=Unordered
Mutation:
=
>
Evolve (Heap array
−
>
Heap list)
Event:
Heap list=Full
Heredity based selection:
Workload=Write intensive
Data access=Unordered
Mutation:
=
>
Evolve (Heap list
−
>
Hash table)
Heap array
Heap list
Hash table
storage structure can have multiple mutation rules mapped to it. These muta-
tion rules consist of three information elements: Event, Heredity based selection,
and Mutation. The event identifies, when this mutation rule should be executed.
Different mutation rules can have the same event, but not all of them execute
the mutation. The heredity based selection identifies precisely, when evolution
should occur based on the heredity information gathered for the existing storage
structure. Heredity information means the gathered statistics about the storage
structure, e.g., workload type, data access pattern, previous evolution details,
etc. The mutation defines the actions that should be executed to evolve the stor-
age structure. Example of an evolution path is shown in Table 7. We envision
that common DBMS maintenance best practices can be documented using the
evolution path mechanism. ECOS assumes that DBMS vendors provide the evo-
lution paths that best suit their DBMS internals, with the provision of alteration
for a database administrator. The only liability for configuration that lies with
database designers and administrator is to have a look at the evolution path
for the DBMS and if needed, alter it with desired changes. Evolution process in
