| |
 |
| Paper: |
Using Containers to Deploy Services on Arbitrary External Infrastructure |
| Volume: |
522, Astronomical Data Analysis Software and Systems XXVII |
| Page: |
359 |
| Authors: |
Major, B.; Ouellette, J. |
| Abstract: |
As the software services offered at the Canadian Astronomy Data Centre (CADC) grow in scale, we increasingly rely on externally provided infrastructure to host these services. This means that we have less control over the hardware and operating systems on which these services are hosted. In order to deal with this very heterogeneous array of infrastructure, we have been able to use Docker containers to define a common ground. If we set the minimal requirement of access to Docker, we are able to confidently build, test, and deploy our services, even if they contain legacy software which is tied to specific hardware or operating systems. However, adopting this type of contract with infrastructure providers creates other new challenges. Software developers are inherently more involved with operational concerns and are thus obliged to take on a 'devops' role. Similarly, the workflow used to test and deploy services must be adapted to fit what is provided by the container paradigm. Also, precautions must be taken with any credentials required by the services when operating on external systems to which the operating partners may have access. We discuss this variety of benefits and challenges faced when moving to a production environment based on a Docker container interface. |
|
|
|
|
|
