Spatial Index Schemes for Cloud Environments - Geographical Information Systems: Trends and Technologies - page 67

Table 7. An example of check-in records.

cid rid rest.name rest.lat rest.lng uid Time

1 p 1 Friday 24.805 120.995 u 1 2011/05/08

2 p 1 Friday 24.805 120.995 u 2 2011/08/08

3 p 2 McDonald's 24.794 121.002 u 2 2011/08/30

4 p 2

McDonald's 24.794 121.002 u 3 2011/10/10

5 p 3 KFC 24.794 121.005 u 4 2011/11/07

Table 8. Data in Cassandra.

(b) A column family for restaurants'
information.

(a) A column family for the simplifi ed
c heck-in information.

columns columns

keys keys

name value name value

rid p 1 name Friday

1 uid u 1 p 1 lat 24.805

time 2011/05/08 lng 120.995

rid p 1 name McDonald's

2 uid u 2 p 2 lat 24.794

time 2011/08/08 lng 121.002

rid p 2 name KFC

3 uid u 2 p 3 lat 24.794

time 2011/08/30 lng 121.005

p 2

rid

4 uid u 3

time 2011/10/10

rid p 3

5 uid u 4

time 2011/11/07

family is illustrated in Table 9. Similarly, in Cassandra, the keys, columns'
names, and super column family names are stored in lexicographical
order. In addition, for HBase and Cassandra, the data of columns are
distributed on servers. The difference between Cassandra and HBase is
that Cassandra allows a column or a super column to be added arbitrarily,
but a column family in HBase cannot be added arbitrarily after a table has
been created.
The data model of RDBMSs are different to the data model of HBase/
Cassandra. The data stored in RDBMSs are structured in tables, fi elds and
records. Specifi cally, each table consists of records and each record consists
of one or more fi elds. Because RDBMSs guarantee the ACID properties, i.e.,
atomicity, consistency, isolation, and durability, RDBMSs are not scalable to
support large data. For instance, if there are multiple records to be updated
in a single transaction, multiple tables will be locked for modifi cation. If