Cloudware Basics - Guide to Cloud Computing for Business and Technology Managers - page 277

infrastructure availability ensuring redundancy in infrastructure, such as
servers, so that new servers are always ready to replace failed servers; mid-
dleware availability achieved with middleware redundancy; and application
availability achieved via application redundancy.

To the 9s (measures of application availability): Service-level
agreements (SLAs) on availability are often measured in 9s. This
describes the target percent of unplanned availability to be
achieved, typically on a monthly or annual basis. Each 9 corre-
sponds to a 10-fold decrease in the amount of downtime. For an impor-
tant application, such as e-mail or a CRM system, three 9s might be a
reasonable target, whereas critical services such as public utilities
would tend to target five 9s.
The following table describes the amount of acceptable downtime
per year for the corresponding level of availability:

# of 9s SLA Target (%) Maximum Downtime per Year

2 99 3 days, 15 h, and 40 min

3 99.9 8 h and 46 min

4 99.99 52 min and 36 s

5 99.999 5 min and 16 s

6 99.9999 31.56 s

The other approach is to build support for high availability into the cloud
infrastructure, which is of two types:

1. Failure detection , where the cloud infrastructure detects failed appli-
cation instances and avoids routing requests to such instances

2. Application recovery , where failed instances of application are
restarted

13.6.3.1 Failure Detection

Many cloud providers, such as Amazon Web Services' Elastic Beanstalk,
detect when an application instance fails and avoid sending new requests to
the failed instance. In order to detect failures, one needs to monitor for failures.

13.6.3.1.1 Failure Monitoring
There are two techniques of failure monitoring. The first method is heart-
beats, where each application instance periodically sends a signal (called
a heartbeat) to a monitoring service in the cloud. If the monitoring service