Databases Reference
In-Depth Information
Quality trees are one method of using graphical presentations to quickly explain the
process and findings of your project to stakeholders. Sometimes the right image or
metaphor will make all the difference. Let's see how finding the right metaphor can
make all the difference in the adoption of new technologies.
12.5.2
Apply your knowledge
Sally is working on a team that has looked at two options to scale out their database.
The company provides data services for approximately 100 large corporate customers
who run reports on only their own data. Only a small number of internal reports span
multiple customers' files. The team is charged with coming up with an architecture
that can scale out and allow the customer base to continue to grow without impacting
database performance.
The team agrees on two options for distributing the customer data over a cluster of
servers. Option A puts all customer data on a single server. This architecture allows for
scale-out and new database servers can be added as the customer base grows. When
new servers are added, they'll use the first letter of the customer's name to divide the
data between the servers. Option B distributes customer records evenly across multi-
ple servers using hashing. Any single customer will have their data on every node in
the cluster and replicated three times for high availability.
Sally's team is concerned that some stakeholders might not understand the trade-
offs and believe that storing all customer data on a single node with replication might
be a better option. But if they implement option A, half of the nodes in the cluster will
be idle and only in use when a failure occurs.
The team wants to show everyone how evenly distributing the queries over the clus-
ter will return faster results even if queries are running on each node in the cluster.
Yet, after a long discussion, the team can't convince their stakeholders of the advan-
tages of their plan. One of the team members has to leave early. He has to catch a
plane and is worried about the long lines at the airport. Suddenly, Sally has an idea.
At their next meeting Sally presents the drawings in figure 12.7.
The next time the group meets, Sally uses her airport metaphor to describe how
most of the nodes in the cluster would remain idle, even when a customer is waiting
many hours for long reports to run. By evenly distributing the data over the cluster,
the load will be evenly distributed and reports will run on every node until the report
finishes. No long lines!
Sally's metaphor works and everyone on the project understands the consequences
of uneven distribution of data in a cluster and how costs can be lower and perfor-
mance improved if the right architecture is used. Though you can create a detailed
report of why one scale-out architecture works better than another using bulleted lists,
graphics, models, or charts, using a metaphor everyone understands gets to the heart
of matter quickly and drives the decision-making process.
Like buying a car, there will seldom be a perfect match for all your needs. Every
year the number of specialized databases seems to grow. Not many free and stan-
dards-compliant open source software systems scale well and store all data types
