Support customizable cpu, memory, network task complexities #47
Conversation
alekodu
changed the title
| "namespace": "default", | ||
| "node": "node-1" | ||
| "cluster": "cluster1", | ||
| "namespace": "edge-namespace", |
There was a problem hiding this comment.
It is suggested to have the existing namespace for all cases, which is default.
| version: cluster1 | ||
| namespace: edge-namespace | ||
| data: | ||
| conf.json: '{"processes":2,"threads":2,"endpoints":[{"name":"endpoint1","protocol":"http","execution_mode":"sequential","cpu_complexity":{"execution_time":"","method":"","workers":0,"cpu_affinity":null,"cpu_load":""},"memory_complexity":{"execution_time":"","method":"","workers":0,"bytes_load":""},"network_complexity":{"forward_requests":"asynchronous","response_payload_size":512,"called_services":[{"service":"service2","port":"80","endpoint":"endpoint1","protocol":"http","traffic_forward_ratio":1,"request_payload_size":256},{"service":"service2","port":"80","endpoint":"endpoint2","protocol":"http","traffic_forward_ratio":1,"request_payload_size":256}]}},{"name":"endpoint2","protocol":"http","execution_mode":"parallel","cpu_complexity":{"execution_time":"","method":"","workers":0,"cpu_affinity":null,"cpu_load":""},"memory_complexity":{"execution_time":"","method":"","workers":0,"bytes_load":""},"network_complexity":{"forward_requests":"asynchronous","response_payload_size":512,"called_services":[]}}]}' |
There was a problem hiding this comment.
Is cpu_affinity true for this example?
There was a problem hiding this comment.
right, I didn't update the example yaml files
| version: cluster1 | ||
| namespace: edge-namespace | ||
| data: | ||
| conf.json: '{"processes":2,"threads":2,"endpoints":[{"name":"endpoint1","protocol":"http","execution_mode":"parallel","cpu_complexity":{"execution_time":"5s","method":"fibonacci","workers":2,"cpu_affinity":[22,23],"cpu_load":"100%"},"memory_complexity":{"execution_time":"5s","method":"swap","workers":24,"bytes_load":"100%"},"network_complexity":{"forward_requests":"asynchronous","response_payload_size":512,"called_services":[]}},{"name":"endpoint2","protocol":"http","execution_mode":"parallel","cpu_complexity":{"execution_time":"","method":"","workers":0,"cpu_affinity":null,"cpu_load":""},"memory_complexity":{"execution_time":"","method":"","workers":0,"bytes_load":""},"network_complexity":{"forward_requests":"asynchronous","response_payload_size":512,"called_services":[]}}]}' |
There was a problem hiding this comment.
Not everyone has access to servers with 24 cores of CPU. I would suggest changing the values of cpu_affinity.
generator/input/new_description.json
Outdated
| ] | ||
| "execution_mode": "sequential", | ||
| "cpu_complexity": { | ||
| "execution_time": "10s", |
There was a problem hiding this comment.
As discussed before, we should reduce the execution_times in the scale of miliseconds per requests.
There was a problem hiding this comment.
It seems the minimum timeout supported by stress-ng is 1 second...
There was a problem hiding this comment.
Is it possible to kill the stress-ng subprocess in python? so we just pass a huge timeout to stress-ng and then kill it will python
There was a problem hiding this comment.
It could be but the load stress-ng generates takes some time to be applied and reflected on resource consumption. If we want to have minimum processing delay in the order of milliseconds we could instead simulate a sleep task or just don't configure any cpu or memory task...
There was a problem hiding this comment.
sleep task does not create any load. I'll check it out.
There was a problem hiding this comment.
yes, it doesn't create any load... I think it is hard to simulate tasks that are both cpu-bounded and I/O bounded at the same time... either we can simulate tasks that have a heavy load on CPU/memory or on network, but not both...
There was a problem hiding this comment.
After discussion, this will be solved in issue #48.
| EpExecModeDefault = "sequential" | ||
| EpNwResponseSizeDefault = 512 | ||
|
|
||
| EpExecTimeDefault = "1s" |
There was a problem hiding this comment.
The default values should be also reduced to the a reasonable scale for a single request.
model/Dockerfile
Outdated
| @@ -20,6 +20,9 @@ RUN mkdir -p /usr/src/app | |||
| RUN apt update | |||
| RUN apt install -y jq \ | |||
| wget \ | |||
| curl \ | |||
| vim \ | |||
There was a problem hiding this comment.
Are we really using vim and curl in the container? I guess we should delete these two from the docker image.
| @@ -43,105 +43,233 @@ def getForwardHeaders(request): | |||
| return headers | |||
|
|
|||
|
|
|||
| def run_task(service_endpoint): | |||
| def run_task(service_name, service_endpoint): | |||
| headers = getForwardHeaders(request) | |||
There was a problem hiding this comment.
We are already have the same value for headers as defined here, also we are importing headers from wsgiref, don't they conflict with each other? And do we really need wsgiref?
There was a problem hiding this comment.
I didn't make changes to the header part... If I remember correctly, this was done to forward jaeger tracing headers through the set of microservices
| RequestPayloadSize int `json:"request_payload_size"` | ||
| } | ||
|
|
||
| type CpuComplexity struct { |
There was a problem hiding this comment.
We need to define optional values for the required values here, so that if the user skip this values, we could generate reasonable yaml files.
closes #14, closes #9, closes #4, closes #38, closes #20
New endpoint format:
"endpoints": [
{
"name": "end1",
"protocol": "http",
"execution_mode": "parallel",
"cpu_complexity": {
"execution_time": "5s",
"method": "fibonacci",
"workers": 2,
"cpu_affinity": [
0,
1
],
"cpu_load": "100%"
},
"memory_complexity": {
"execution_time": "5s",
"method": "swap",
"workers": 24,
"bytes_load": "100%"
},
"network_complexity": {
"forward_requests": "asynchronous",
"response_payload_size": 512,
"called_services": [
{
"service": "service2",
"port": "80",
"endpoint": "end2",
"protocol": "http",
"traffic_forward_ratio": 1,
"request_payload_size": 256
}
]
}
}
]
New response format:
{
"cpu_task":{
"services":[
"service1/end1", "service2/end2"
],
"statuses":[
"stress-ng: info: [76] dispatching hogs: 2 cpu\nstress-ng: info: [76] successful run completed in 5.00s\nstress-ng: info: [76] stressor bogo ops real time usr time sys time bogo ops/s bogo ops/s\nstress-ng: info: [76] (secs) (secs) (secs) (real time) (usr+sys time)\nstress-ng: info: [76] cpu 64633075 5.00 9.61 0.00 12926973.13 6725606.14\n",
"stress-ng: info: [33] dispatching hogs: 2 cpu\nstress-ng: info: [33] successful run completed in 5.32s\nstress-ng: info: [33] stressor bogo ops real time usr time sys time bogo ops/s bogo ops/s\nstress-ng: info: [33] (secs) (secs) (secs) (real time) (usr+sys time)\nstress-ng: info: [33] cpu 4349696 5.17 0.80 0.00 841800.59 5437120.00\n"
]
},
"memory_task":{
"services":[
"service1/end1", "service2/end2"
],
"statuses":[
"stress-ng: info: [80] dispatching hogs: 24 vm\nstress-ng: info: [80] successful run completed in 5.28s\nstress-ng: info: [80] stressor bogo ops real time usr time sys time bogo ops/s bogo ops/s\nstress-ng: info: [80] (secs) (secs) (secs) (real time) (usr+sys time)\nstress-ng: info: [80] vm 0 5.23 4.82 5.43 0.00 0.00\n",
"stress-ng: info: [35] dispatching hogs: 24 vm\nstress-ng: info: [35] successful run completed in 5.33s\nstress-ng: info: [35] stressor bogo ops real time usr time sys time bogo ops/s bogo ops/s\nstress-ng: info: [35] (secs) (secs) (secs) (real time) (usr+sys time)\nstress-ng: info: [35] vm 0 5.31 4.03 5.62 0.00 0.00\n"
]
},
"network_task":{
"services":[
"(service1/end1, service2/end2)"
],
"statuses":[
200
],
"payload":"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"
}
}