Tried and Trusted

Salt Can Work with All Clouds

Salt is the only application considered here that passes individual API operations directly to the system administrator's command line without messing up the application's internal records. Therefore, on a running virtual machine (VM), you can remove a volume from Salt via the AWS API, and Salt will remember the change and the new state of the resource in AWS. Additionally, all standard API calls for EC2 can also be run by Salt because the vendor maintains the EC2 integration. Conveniently, when you use Salt to create a VM in AWS, the Salt agent is also installed.

The support for Azure and GCP is practically identical. Clearly, the manufacturer maintains a list of basic features internally and consistently supports those features for every cloud environment. Of use are Python development kits for Azure, AWS, and the like that provide the lion's share of the required functionality. In keeping with the spirit of genuine FL/OSS software, Salt avoids reinventing the wheel in terms of its integration of Azure and others; instead, it relies on the components that exist there, which not only avoids duplicate code in Salt, but also indirectly means better Python code because Salt returns error fixes to the developer.

Private Clouds with Salt

If you want to unleash Salt on a private cloud, you will find excellent support for OpenStack in the software. Because the OpenStack APIs are open and well documented, it is easy for providers to integrate appropriate support into their own products (Figure 3).

Figure 3: Thanks to extensive community work, all of the tools (e.g., Ansible, as shown here) also support OpenStack as a target platform.

That is exactly what the Salt developers have done with regard to OpenStack. Instances and volumes can be created and deleted just as easily as images. Apparently even less relevant special functions, such as the interfaces exposed by SSH, can be controlled by Salt.

Conclusions

Ansible, Chef, Puppet, and Salt impressively demonstrate that you do not always need a proprietary tool if you want to start and manage workloads in clouds. A few combinations stand out from the crowd: Puppet and AWS have recognizably joined forces and even offer product bundles that clearly outperform normal automation in terms of the available feature set. Of course, admins will need deep pockets to benefit. Chef, Ansible, and Salt are a little more casual and offer products that are similar to Puppet, but they don't push them as hard commercially.

Indirectly, however, this also means that anyone who is still deciding on a suitable automation provider for their own environment should at least factor in their intended cloud use. The candidates in this test go about their duties impeccably. In detail, however, some differences can make one of the solutions appear more or less suitable for an individual application. Finally, because all of the solutions used are based on open source software, extensive testing is possible before making a decision – and it is strongly recommended.