Tweaked vSwarm deployer

[aryans1204]

Summary

A small summary of the requirements (in one/two sentences).

Implementation Notes ⚒️

• Briefly outline the overall technical solution. If necessary, identify talking points where the reviewer's attention should be drawn to.
  Goal of this PR is to add functionality in the loader to be able to execute vSwarm functions from a mapper_output.json file.

The added functionality pertains to:

1. Added parsing capabilities under pkg/trace/mapper_trace_parser.go to parse mapper_output.json files and extract vSwarm functions.
2. Added deployment capability under pkg/driver/deployment/knative.go to correctly deploy vSwarm functions, and a yamls.tar.gz tarball that contains the YAML files for all the vSwarm functions. Documentation about this tarball is embedded inside the tarball.

This extends upon the work done previously on the invoker, to be able to execute vSwarm functions. With the merge of this PR, vSwarm functions will be able to be executed from the loader, by specifying config options.

Added parsing, and deployment functionality to run vSwarm functions using Invitro loader, based on mapper output.

External Dependencies 🍀

Breaking API Changes ⚠️

Simply specify none (N/A) if not applicable.

[leokondrashov]

Please explain in details why and what are you doing in the PR description. The template is there for a reason.

I don't see any documentation in the changes as well. There should be something about the feature, file formats (you are introducing several).

[leokondrashov]

Please add docs on all files that you add as inputs to the execution. I have no idea how to get deploy_info and deployment yamls for a new set of benchmarks. What are the requirements for those benchmarks (use the relay from vSwarm, at least). Where is each file should be placed to be properly found by the loader.

[leokondrashov]

Can you explain how I should create this file? Does something generate it? Because if my memory doesn't lie to me, previously, there were redeployment commands, not paths. Is there anything else that depends on the file formatted that way?

[aryans1204]

So the deploy_info.json file exists inside the yamls.tar.gz tarball, where the file is generated using the generate_deploy_info.py file, also inside the tarball. Further, the docs to generate deploy info and its purpose are also inside the tarball. However, in order to generate the YAMLs, I just copied over the YAMLs from vSwarm into here, with some minor changes like changing the service name to the environment variable FUNC_NAME etc. In the future if someone needs to execute new benchmarks, they will have to generate the YAMLs and their locations by themself. I can add documentation for this in the tarball as well. Since the scope of this PR is to support vSwarm functions to be executed from a given mapper_output file, using the invitro loader, and not to generically execute any benchmark.

[leokondrashov]

The docs inside the tarball is an extremely bad idea. Don't put it inside.

I want to see an explanation of each field used by the loader and how it affects the end result. Since you introduce a new deployment procedure, you should explain it here. If you have the script for that, explain it as well. If you expect someone else to do something, explain what needs to be done and what the requirements are for the outcome of these actions. With these docs, it should be possible to use the loader to run new benchmarks without looking into the code. It indeed doesn't need to explain all of the steps, but the user interface (file formats, effects of each field, etc)

[aryans1204]

Currently, there are docs under docs/loader.md which explain all the steps required for a user to execute vSwarm benchmarks, ie. untar the YAML tarball, and then the execution command for executing under vSwarm mode. These docs are present under the Executing vSwarm functions and Single load generation subheadings of the loader docs. Please let me know whether they are adequate or some aspects are still unclear to the user about vSwarm based execution. As for the fields added, in terms of the experiment config, there is only one flag, ie. the vSwarm flag which is needed to turn on vSwarm execution. This documentation for the field is also added under docs/configuration.md. As for the deploy_info docs, I will move that doc from the tarball to the docs/ subdir. In terms of the changes to deployment procedure, there is an addition of a new script for deploying the databases for some vSwarm benchmarks. What would be a good way to create documentation for this? Under loader.md or a separate file under docs/?

[JooyoungPark73]

I think docs/generate_deploy_info_docs.md is okay. Please add schema and description of deploy_info.json

Also, pull all the Python scripts out of tarball and move to tools/mapper or sth, as it has to be transparent and can be viewed from Github webpage. Also, your tarball includes json file, so rename tarball to something more general.