09.02
High availability can be analyzed from various aspects and angles of the system as a unit and depends directly on the application design, architecture, hardware tools and not to forget expertise available for maintenance. To achieve a high level of availability every aspect plays an important role and must be analyzed and measured.
Most systems do use one or another flavor of database behind the scene which is always considered as the backbone of information storage and retrieval. The most common databases used today commonly are Oracle, MS-SQL, MySQL and db2. Regardless the type of database model the question remains “How soon the resource can be replaced and normal operation be resumed without any or minimum loss of data?”.
To define the Availability we must consider the basic metric which is an availability of application (system) or non-accessibility of the application i.e. amount of downtime. Availability is usually measured over a course of time generally a year or even month. The period must be defined before any anticipations can be presented. When we have defined a period, then comes the question how to measure the availability or achieve a particular level of availability. The most common formula used is
The above formula provides a percentage how often a candidate application is available.
Another common way of calculating the availability is the count of 9’s, i.e. how many 9’s are present in the percentage of availability. For example three 9’s means 99.9% availability of the application over the course of the period defined and whatever is left over is considered as an acceptable downtime.The following chart describes the level of availability and downtime in terms of count of 9’s
Count of 9’s Availability Chart
| Availability | count of 9’s | Downtime |
|---|---|---|
| 90% | 1 nine | 36.5 days per year |
| 99% | 2 nine | 3.65 days per year |
| 99.9% | 3 nine | 8.76 hours per year |
| 99.99% | 4 nine | 52 minutes per year |
| 99.999% | 5 nine | 5 minutes per year |
| 99.9999% | 6 nine | 31 seconds per year |
As we can see the number of 9’s increase the amount of downtime decreases but the decrease rate of downtime is slower. For example from 2 to 3 nines reduced the amount of downtime by approx 2 days where as from 5 to 6 it only in minutes. One more things starts happening when the number of 9’s increase the cost of maintaining the availability increases and if the number resources that make up the application are more the share of downtime each resource is allowed also gets distributed which leads to each resource or component to be redundant have to offer fast fail-safe architecture. Each resource also adds up to maintenance expenses.
Now the analysts must take into consideration all these factors before deciding what level of availability the system should offer. Another factor which may play a significant role is to layout what time during the day the application need to be available. Whether application serves global customers or is it specific to a region or business hours. Global service application are expected to achieve at least three 9’s.
How many 9’s we need is sometimes hard to predict but certainly a guideline for an application/system availability. Another formula commonly used is as follows
MTBF = Mean time between failures
MTTR = Mean time to repair
The above formula can estimate the future availability and this ratio is supplied by most hardware vendors which allows MTTR to be calculated on various pre-defined procedures in place.
No Comment.
Add Your Comment