|
Thursday, December 16, 16:05
Now that the team had formed data domains from the monolithic Sysops Squad database, Devon noticed that the Survey data domain would be a great candidate for migrating from a traditional relational database to a document database using JSON. However, Dana, the head of data architecture, didn’t agree and wanted to keep the tables as relational.
“I simply don’t agree,” said Dana. “The survey tables have always worked in the past as relational tables, so I see no reason to change things around. "
“Actually,” said Skyler, “if you had originally talked with us about this when the system was first being developed, you would understand that from a user interface perspective, it’s really hard to deal with relational data for something like a customer survey. So I disagree. It may work out good for you, but from a user interface development standpoint, dealing with relational data for the survey stuff has been a major pain point.”
“See, so there you are,” said Devon. “This is why we need to change it to a document database.”
“You seem to forget that as the data architect for this company, I am the one who has ultimate responsibility for all these different databases. You can’t just start adding different database types to the system,” said Dana.
“But it would be a much better solution,” said Devon.
“Sorry, but I’m not going to cause a disruptor on the database teams just so Skyler can have an easier job maintaining the user interface. Things don’t work that way.”
“Wait,” said Skyler, “didn’t we all agree that part of the problem of the current monolithic Sysops Squad application was that the development teams didn’t work close enough with the database teams?”
“Yes,” said Dana.
“Well then,” said Skyler, “let’s do that. Let’s work together to figure this out.”
“OK,” said Dana, “but what I’m going to need from you and Devon is a good solid justification for introducing another type of database into the mix.”
“You got it,” said Devon. “We’ll start working on that right away.”
Devon and Skyler knew that a document database would be a much better solution for the customer survey data, but they weren’t sure how to build the right justifications for Dana to agree to migrate the data. Skyler suggested that they meet with Addison to get some help, since both agreed that this was somewhat an architectural concern. Addison agreed to help, and set up a meeting with Parker (the Sysops Squad product owner) to validate whether there was any business justification to migrating the customer survey tables to a document database.
“Thanks for meeting with us, Parker,” said Addison. “As I mentioned to you before, we are thinking of changing the way the customer survey data is stored, and have a few questions for you.”
“Well,” said Parker, “that was one of the reasons why I agreed to this meeting. You see, the customer survey part of the system has been a major pain point for the marketing department, as well as for me.”
“Huh?” asked Skyler. “What do you mean?”
“How long does it take you to apply even the smallest of change requests to the customer surveys?” asked Parker.
“Well,” said Devon, “it’s not too bad from the database side. I mean, it’s a matter of adding a new column for a new question or changing the answer type.”
“Hold on,” said Skyler. “Sorry, but for me it’s a major change, even when you add an additional question. You have no idea how hard it is to query all of that relational data and render a customer survey in the user interface. So, my answer is, a very long time.”
“Listen,” said Parker. “We on the business side of things get very frustrated ourselves when even the simplest of changes take you literally days to do. It’s simply not acceptable.”
“I think I can help here,” said Addison. “So Parker, what you’re saying is that the customer survey changes frequently, and it is taking too long to make the changes?”
“Correct,” said Parker. “The marketing department not only wants better flexibility in the customer surveys, but better response from the IT department as well. Many times they don’t place change requests because they know it will just end in frustration and additional cost they didn’t plan for.”
“What if I were to tell you that the lack of flexibility and responsiveness to change requests has everything to do with the technology used to store customer surveys, and that by changing the way we store data, we could significantly improve flexibility as well as response time for change requests?” asked Addison.
“Then I would be the happiest person on Earth, as would the marketing department,” said Parker.
“Devon and Skyler, I think we have our business justification,” said Addison.
With the business justification established, Devon, Skyler, and Addison convinced Dana to use a document database. Now the team had to figure out the optimal structure for the customer survey data. The existing relational database tables are illustrated in Figure 6-34. Each customer survey consisted of two primary tables—a Survey table and a Question table, with a one-to-many relationship between the two tables.
Figure 6-34. Tables and relationships in the sysops survey data domain
An example of the data contained in each table is shown Figure 6-35, where the Question table contains the question, the answer options, and the data type for the answer.
Figure 6-35. Relational data in tables for survey and question in the survey data domain
“So, essentially we have two options for modeling the survey questions in a document database,” said Devon. “A single aggregate document or one that is split.”
“How to we know which one to use?” asked Skyler, happy that the development teams were now finally working with the database teams to arrive at a unified solution.
“I know,” said Addison, “let’s model both so we can visually see the trade-offs with each approach.”
Devon showed the team that with the single aggregate option, as shown in Figure 6-36, with the corresponding source code listing in Example 6-3, both the survey data and all related question data were stored as one document. Therefore, the entire customer survey could be retrieved from the database by using a single get operation, making it easy for Skyler and others on the development team to work with the data.
Figure 6-36. Survey model with single aggregate
Example 6-3. JSON document for single aggregate design with children embedded
#
Survey
aggregate
with
embedded
questions
{
Do'stlaringiz bilan baham: |
