Databases Reference
In-Depth Information
P
RECISION
SERVICE
MONITORING
AND
NOTIFICATION
In addition to specifying service levels, you can also invest in tools to monitor your ser-
vice level. When you configure the number of reads per second a service performs, set-
ting the parameter too low may mean the user would experience a delay during peak
times. By using a simple
API
, detailed reports showing the expected versus actual loads
can point you to system bottlenecks that may need additional resource adjustments.
Automatic notification systems can also trigger email messages when the volume of
reads or writes exceeds a threshold within a specified period of time. For example,
you may want to send an email notification if the number of reads per second exceeds
80% of some predefined number within a 30-minute period. The email message
could contain a link to the monitoring tools as well as links that would allow you to
add more servers if the service level was critical to your users.
S
CALABILITY
AND
RELIABILITY
When a database interface is simple, the resulting systems can have higher scalability
and reliability. This means you can tune any solution to the desired requirements.
Keeping an interface simple allows novice as well as advanced data modelers to build
systems that utilize this power. Your only responsibility is to understand how to put this
power to work solving business problems.
A simple interface allows you to focus on load and stress testing and monitoring of
service levels. Because a key-value store is simple to set up, you can spend more time
looking at how long it takes to put or get 10,000 items. It also allows you to share these
load- and stress-testing tools with other mem-
bers of your development team.
Low portability
High portability
Application
Application
P
ORTABILITY
AND
LOWER
OPERATIONAL
COSTS
One of the challenges for information systems
managers is to continually look for ways to
lower their operational costs of deploying sys-
tems. It's unlikely that a single vendor or solu-
tion will have the lowest cost for all of your
business problems. Ideally, information sys-
tems managers would like to annually request
data service bids from their database vendors.
In the traditional relational database world, this
is impractical since porting applications
between systems is too expensive compared to
the relative savings of hosting your data on a
new vendor's system. The more complicated
and nonstandardized they are, the less portable
they can be and the more difficult moving them
to the lowest cost operator is (see figure 4.4).
Database
Database
Complex and
nonstandard APIs
Simple and
standard APIs
Figure 4.4
Portability of any application
depends on database interface complexity.
The low-portability system on the left has
many complex interfaces between the
application and the database, and porting
the application between two databases
might be a complex process requiring a
large testing effort. In contrast, the high-
portability application on the right only
uses a few standardized interfaces, such
as put, get, and delete, and could be
quickly ported to a new database with
lower testing cost.
