Virtualizing Resources for the Cloud - Large Scale and Big Data: Processing and Management - page 529

with local instance-store disks as root devices. Instance-store volumes are volatile
storages and cannot survive stops and terminations. Second, EC2 instances can be
attached to Elastic Block Storages (EBS), which provide raw block devices. The block
devices can then be formatted and mounted with any file system at EC2 instances.
EBS volumes are persistent storages and can survive stops and terminations. EBS
volumes of sizes from 1 GB to 1 TB can be defined and RAID arrays can be created
by combining two or more volumes. EBS volumes can even be attached or detached
from instances while they are running. They can also be moved from one instance
to another, thus rendering independent of any instance. Finally, applications running
on EC2 instances can access Amazon S3 through a defined API. Amazon S3 is a
storage that makes web-scale computing easier for developers, whereby any amount
of data can be stored and retrieved at any time and from anywhere on the web [5].
To this end, Amazon EC2 users do not only have the flexibility of choosing among
many instance and storage types, but further have the capability of mapping elastic
IP addresses to EC2 instances, without a network administrator's help or the need to
wait for DNS to propagate new bindings. Elastic IP addresses are static IP addresses,
but tailored for the dynamicity of the cloud. For example, unlike a traditional static
IP address, an Elastic IP address enables tolerating an instance failure by program-
matically remapping the address to any other healthy instance under the same user
account. Thus, Elastic IP addresses are associated with user accounts and not EC2
instances. Elastic IP addresses exist until explicitly removed and persist even while
accounts have no current running instances.

16.9.2.4 Fault Tolerance
Amazon EC2 users are capable of placing instances and storing data at multiple
locations represented as regions and availability zones . As shown in Figure 16.25,
a region can consist of one or many availability zones and an availability zone can
consist of typically many blade servers. Regions are independent collections of
AWS resources that are geographically dispersed to avoid catastrophic disasters. An

Region A Region B

Availability
zone A

Availability
zone B

Availability
zone A

Availability
zone B

Availability
zone C

FIGURE 16.25 Regions and availability zones in AWS cloud platform. Regions are geo-
graphically dispersed to avoid disasters. Availability zones are engineered as autonomous
failure zones within regions.