Create README for KubeCon keynote comparison

lixuna commented 5 years ago

Create README for KubeCon keynote comparison

[x] Create https://github.com/cncf/cnfs/tree/master/comparison/kubecon18-chained_nf_test/README.md
[x] Short description (one line: ie “Comparing performance of NFs on OpenStack and K8s”)
[ ] Add an aspirational message of test bed (ie. Ideally, you'll only need an API key from Packet/bare metal to deploy and replicate results. Currently, some customization is needed on machines and BIOS configuration, etc.)
[ ] What is this test comparison? (Add a description of test comparison)
[ ] Where are we right now? (compared to aspiration)
[x] Document the requirements for the workstation that does the deployment and runs the tests
[ ] Describe the K8s and OpenStack testbed
- [ ] Add information about systems being tested
  - See https://github.com/cncf/cnfs/issues/117
- [x] Add wiring diagrams showing machines, switches and layer-2 network
[ ] Describe the test cases (eg. snake and pipeline)
- add high-level diagram or link to another document. See https://github.com/cncf/cnfs/issues/155
[ ] Test results
- [ ] Add summary of test results
- [ ] Add links to more extensive results or nfvbench logs

Steps to recreate test results

Add steps to recreate test results, including manual steps for configuring reserved instances and BIOS tasks

Common

[ ] Packet account
[ ] Pre-req software to install

Reserved instances

[ ] BIOS customizations
[ ] Hardware setup information

TBD

On-Demand instances

TBD

taylor commented 5 years ago

Audience: Person who wants to recreate the results themselves on Packet dual mellanox machines

High level requirements

Packet: account, Project ID, API key, SSH key
End-user workstation: ssh, git, docker
- Recommended: Modern Linux OS

Pre-req steps:

Install GIT
Install docker

General steps:

git clone git@github.com:cncf/cnfs.git
cd cnfs/comparison/kubecon18-chained_nf_test
# create environment .env with Packet, SSH and test case information
# load .env
. .env
./deploy_k8s_test_case
./run_k8s_test_case # connects to traffic generator, runs tests, collects result & shows summarized results on console

Audience: Internal dev

What we have right now

K8s cluster

# create environment .env with Packet, SSH and test case information
# load .env
. .env

./deploy_k8s_cluster   # using cross-cloud - terraform, cloud-init
./setup_k8s_l2   # sets up l2 for Packet switch and worker nodes
./deploy_cnfs  # using helm charts

Traffic generator

# create environment .env with Packet, SSH and test case information
# load .env
. .env
./deploy_traffic_generator  # terraform + ansible from the docker container

Run k8s test with traffic generator

./run_k8s_testcase  #  ansible running nfvbench on the traffic generator

Similar steps for OpenStack

Breaking down some of the steps

deploy_k8s_cluster

builds the tool container out of the repo
runs the tool container
- we need to set the environment configuration which passed to docker
runs terraform from in the container
runs ansible from the in the container

denverwilliams commented 5 years ago

Steps for the Packet Generator

git clone https://github.com/cncf/cnfs.git
cd cnfs/comparison/cnf_edge_throughput/packet_generator
export PACKET_PROJECT_ID=YOUR_PACKET_PROJECT_ID 
export PACKET_AUTH_TOKEN=YOUR_PACKET_API_KEY
export PACKET_FACILITY="sjc1"
./deploy_packet_generator.sh dual_mellanox

Full Readme is here https://github.com/cncf/cnfs/tree/master/comparison/kubecon18-chained_nf_test/packet_generator

taylor commented 5 years ago

We can create additional steps for the quad intel x710 machines.

Long term, maybe not for Kubecon

We can also, have steps for performance tuning that require reserved instances. This will include pre-reqs for setup of the reserved instances and BIOS configuration.

Alternative, would be adding support for pre-created machines and do not delete them. k8s would have to provision existing instances.

taylor commented 5 years ago

Add a prominent notice to the README that the end-user is responsible for all Packet charges and updates in the project chosen. (eg. instances deleted or left around)