Optimizing Aggregate Query Processing in Cloud Data Warehouses - Data Management in Cloud, Grid and P2P Systems - page 4

Database Reference

In-Depth Information

3 Our Approach

During query execution, a query accesses data from different tables by performing

join operation on primary and foreign key attributes of those tables, and then the

required attribute values are filtered. This becomes more complicated in the vir-

tual cloud environment, where the location of the data changes more often based

on the resource availability or SLA (Service Level Agreement) of the customer. In-

creased node-to-node communication results in increased network delay and pro-

cessing time, leading to the decrease in query performance. Hence, we need storage

structures that help to minimize the communication between machines, minimize

the maintenance during updates and minimize the increase in the storage space.

3.1 Storage Structure

PK-map Structure: We create a PK-map (i.e., Primary Key map) for each of

the primary keys in the tables of the star schema. A PK-map will have one col-

umn for the primary key and one column for each of the foreign keys referencing

this primary key as shown in Table 1. Primary key column contains primary key

values and foreign key column contains logical record-ids. These logical record-

ids are index to the Tuple-index-map of the foreign key. Tuple-index-map is

explained in the next subsection. The PK-maps are sorted on primary key val-

ues, which allow us to apply run length encoding on foreign key logical record-ids

and reduce the size of the map to a great extent. Thus, the overall size of the

map will be proportional to the number of records in the dimension table, which

are usually smaller in size than fact table.

Tuple-index-map Structure: We create a Tuple-index-map for each of the

foreign key attribute, which reference the primary key attribute of the dimension

table as shown in Table 2. This Tuple-index-map will store the mapping between

the logical and actual record-id of the foreign keys in the foreign key table.

Tabl e 1. NationKey-map and SuppKey-map

Tabl e 2. SuppKey Tuple-index-

map

PS.sk index PS.sk t index

Row 0

N.nationkey S.nk index C.nk index

Row 0

0000000000

0

0

0

Row 1

0000000001

4

6

......

24

Partition 1

Row 2

0000000002

14

26

......

.....

..

..........

..........

..........

Row 400000

.....

..

..........

..........

..........

Row 400001

.....

Partition 2

Row 24 0000000024

..........

..........

......

.....

.....

......

.....

.....

......

.....

.....

S.suppkey PS.sk index

......

.....

.....

Row 0

0000000000

0

......

.....

.....

.........

..........

..........

Partition 1

......

.....

.....

Row 20000 0000020000

100000

......

.....

Partition 49

Row 20001 0000020001

..........

Partition 2

Row 19600000

.....

.........

..........

..........

......

.....

.........

..........

..........

..........

......

.....

Partition 50

.........

..........

..........

Partition 49

......

.....

Row 980001 0000980001

..........

Row 20000000

.....

...........

..........

..........

Partition 50

Row 1000000 0001000000

..........

Next Page

Data Management in Cloud, Grid and P2P Systems

Search WWH ::

Custom Search

Home