INTRODUCTION
When it comes to enterprise software delivery, no single tool or product suite can support the
specialized work of all the varied teams that develop and support a product feature or service. All
these specialists who help plan, build and deliver software require their own role-specific system
to work in.
Integration, therefore, plays a critical role in tying all these tools together and making them work
like a single well-oiled machine. Information flows automatically from tool to tool, specialist to
specialist, “pushing work along” to the next step in the process.
Integration also enables organizations to aggregate data across lines of business or departments,
creating visibility for the managers and executives who need to see the big picture of how IT and
software delivery generate revenue for the business.
The main reason being that the traditional approach to integration – i.e., point-to-point
integration between two tools - cannot handle the size and complexity of the work that
enterprises are undertaking.
The average enterprise and agency needs to flow rich product lifecycle data across a
software delivery value stream that comprises:*
• 100-1000s of projects • With 30-100 fields each
• In 5-10 core tools • And 100s of possible values and states
• Housing 30-40 artifact types
Crucially, any integration requires 'field mapping'. Field mapping defines the shape the artifact should take
on when it flows to a different tool, namely the mapping between fields and values in both repositories.
Before we expand on what model-based integration is and why it’s far superior to point-to-point, it helps to
understand the limitations and fragility of a single point-to-point mapping.
Stakeholders need to decide what data to flow and under which circumstances, in a process that usually
takes significant calendar time. You’ll need to decide which fields are worth synchronizing, and then map
them to each other individually, e.g. 'Name' after maps to 'Summary', 'Description' maps to 'Details', 'Priority'
maps to 'Severity', 'Comments' maps to 'Notes', 'Resolution' maps to a Customer field, and so on.
Then, for each possible field value in one repository, you will need to determine the equivalent value in the
other repository. For example, you’ll need to define how 'Priority' values "1-5" map to 'Severity' values "high-
low". You’ll also have to decide how a specific status in one repository should affect the status of an artifact
in the other repository. For example “ready for test” in an Agile planning tool will set the status to “New” in a
test management tool.
Once determined, it can take approximately three hours to implement the desired behavior and field
mapping in the integration of a single project pair. The process must then be repeated for each project pair,
for each artifact type.
When integrating ten projects with two artifact types across a single pair of repositories, you can be
looking at more than 50 hours of point-to-point mapping. For three repositories and four artifact types,
those same ten projects can take more than 200 hours to integrate.
Perhaps once a year you need to add a field to an artifact, e.g. "Service Release", or change a workflow (new
status value "To Triage"). Or perhaps you have a more frequent field value update to make (adding a new
release value to a dropdown list every two weeks).
These types of updates will require amending the field mapping for each project pair and artifact type,
generating approximately 400 hours of maintenance effort per year for those same ten projects.
Scaling point-to-point mapping
For each new product the business wants, organizations typically
create new projects in their tools, which means periodically you’ll
need to add new projects to your existing integrations.
Data consistency
When you’re configuring tens of mappings between individual
project pairs, it’s extremely hard to maintain consistency. Even
with the best of intentions, you’ll forget what you’ve already done
and do it slightly differently each time. Alternatively, you may find
yourself toggling back and forth between an old and new setup and
making small copy/paste mistakes. Most likely, it will be someone
else entirely doing this new configuration, who won’t have the
previous context and will apply their own perspective.
Model-based Integration:
A Revolution
Model-based integration is a fundamentally different approach to
integration that meets the needs of Enterprise IT.
Models are defined upfront and in advance on the integration platform to prevent ad-hoc, subjective or
arbitrary data selection and field mapping. As data flows between two repositories, it goes through the
model. A good analogy to help understand how models work is the act of translating between people who
speak different languages.
If you have two people that speak two different languages, you need to translate only between those
two points. If, however, you have three different languages, you have three points of disconnect in
communication that need to be translated. If you have six, you have 15, if you have 10, you have 45
and so on.
As you can see, resolving these point-to-point disconnects individually quickly becomes unsustainable
given the sheer number of them that can arise. This is how models save the day, acting as a “universal
translator,” overcoming all of the communication disconnects that are present by translating between all
of the points at once.
The beauty of models is that they do not force change on the software delivery teams, nor disrupt their
work. Instead of forcing everyone to use the same tools, processes and systems, models guarantee agility
and autonomy to the teams.
These huge savings translate into faster time to value (TtV) for your products; integration setup will no longer
be a hindrance to project kickoff, and small changes to running integrations can be done in a matter of
minutes. In addition, you don’t need a team of admins to do the work - a single admin should suffice.
To further understand how models eliminate the overhead of point-to-point mapping, let’s take an example
of four projects in Repository A that need to be mapped to four projects in Repository B.
A B
A B
A1 B1
A2 B2
A1 A2 B1 B2
A3 B3
A3 A4 B3 B4
B4 Defect
A4
Model
Now, imagine you’ve recently added two projects, representing two new products in your systems.
A B
A B
A1 B1
A2 B2
A1 A2 B1 B2
A3 B3
A3 A4 B3 B4
A4 B4
A5 A6 Defect B5 B6
A5 B5 Model
A6 B6
Models vs. Templates: Overcoming the Insufficiencies of Field Mapping Templates
Some integration solutions try to solve the scalability challenges of point-to-point mapping with field
mapping templates.
A field mapping template, similar to a PowerPoint template, provides a starting point for your field mapping.
However, templates don’t solve the scalability challenges of integration at scale.
Reusable Templates must be defined per repository pair and per artifact
in multiple Yes No type. For example, Jira Story to VSTS Story, ALM Defect to VSTS
integrations? Bug, ALM Defect to Jira Bug.
Normalizes and
Templates do not standardize data as it flows between tools,
optimizes data for Yes No
requiring extensive effort to produce reports.
reporting?
Tasktop Integration Hub is a reliable, scalable and easy-to-use integration platform, predicated on models.
Launched in 2017, Tasktop Integration Hub is the only commercial off-the-shelf solution for software
delivery toolchain integration and value stream management via model-based integration.
• You need to scale to many more teams, projects, tools and artifact types.
• You have sophisticated reporting and value stream management needs.
• You want your integration to go beyond synchronization to different integration styles, for example
event-driven artifact creation or update based on events from the CI/CD pipeline.
For long-term, strategic integration, models are the only viable option.
A model-based design also requires that the integration platform behavior, too, be standardized. Because the
platform will communicate with tens or hundreds of best-of-breed tools, it must ensure that each of those
exchanges conform to a strictly defined definition of artifacts.
Tasktop does this through our Integration Factory. The Integration Factory has two parts - specification and
automated testing.
To achieve uniformity, scalability and robustness, Tasktop has invested heavily in an elaborate specification
that all tool connectors must adhere to. Tasktop’s specification has nine high level epics broken down into
detailed stories, focusing on how we support connectivity to the repository, artifact types, standard fields,
custom fields, attachments, comments, person mapping, artifact relationships (links).
The Integration Factory also has a robust automated testing capability that can programmatically test all
developed connections against the specification on an ongoing basis. Our test specification has over 1,100
acceptance criteria, and each night it runs over half a million tests on the 57 repositories and 300+ versions
supported by Tasktop Integration Hub.
© 2007-2018 Tasktop Technologies, Inc. All rights reserved. Tasktop and the “Less
is More” logo (<=>) are registered trademarks of Tasktop Technologies, Inc. All other
tasktop.com
trademarks are the property of their respective owners. TSWP-1-2601
@tasktop