Commit 2c4a714a authored by Brewer, Wes's avatar Brewer, Wes
Browse files

Add --scale parameter for scaling telemetry to work on other systems

parent fbde139a

args.py

+1 −0
Original line number Diff line number Diff line
@@ -25,6 +25,7 @@ parser.add_argument('-p', '--plot', nargs='+', choices=['power', 'loss', 'pue', 
                    help='Specify one or more types of plots to generate: power, loss, pue, util, temp')

choices = ['png', 'svg', 'jpg', 'pdf', 'eps']

parser.add_argument('--imtype', type=str, choices=choices, default=choices[0], help='Plot image type')

parser.add_argument('--scale', type=int, default=0, help='Scale telemetry to fit on target system/partition (currently only suupported for marconi100 data)')

parser.add_argument('--system', type=str, default='frontier', help='System config to use')

choices = [policy.value for policy in PolicyType]

parser.add_argument('-s', '--schedule', type=str, choices=choices, default=choices[0], help='Schedule policy to use')

config/setonix-cpu/scheduler.json

+1 −1
Original line number Diff line number Diff line
@@ -3,7 +3,7 @@ 
    "JOB_ARRIVAL_TIME": 900,

    "MTBF": 11,

    "MAX_TIME": 88200,

    "TRACE_QUANTA": 15,

    "TRACE_QUANTA": 20,

    "MIN_WALL_TIME": 3600,

    "MAX_WALL_TIME": 43200,

    "UI_UPDATE_FREQ": 900,

config/setonix-gpu/scheduler.json

+1 −1
Original line number Diff line number Diff line
@@ -3,7 +3,7 @@ 
    "JOB_ARRIVAL_TIME": 900,

    "MTBF": 11,

    "MAX_TIME": 88200,

    "TRACE_QUANTA": 15,

    "TRACE_QUANTA": 20,

    "MIN_WALL_TIME": 3600,

    "MAX_WALL_TIME": 43200,

    "UI_UPDATE_FREQ": 900,

main.py

+2 −1
Original line number Diff line number Diff line
@@ -87,9 +87,10 @@ if args.replay: 
        print(f"Loading {args.replay[0]}...")

        jobs = td.load_snapshot(args.replay[0])

        if args.reschedule:

            print("available nodes:", config['AVAILABLE_NODES'])

            for job in tqdm(jobs, desc="Updating requested_nodes"):

                job['requested_nodes'] = None

                job['submit_time'] = next_arrival()

                job['submit_time'] = next_arrival(1 / config['JOB_ARRIVAL_TIME'])

    else:

        print(*args.replay)

        jobs = td.load_data(args.replay)

raps/dataloaders/marconi100.py

+4 −0
Original line number Diff line number Diff line
@@ -22,6 +22,7 @@ 


"""

import uuid

import random

import pandas as pd

from tqdm import tqdm
@@ -62,6 +63,7 @@ def load_data_from_df(jobs_df: pd.DataFrame, **kwargs): 
    fastforward = kwargs.get('fastforward')

    validate = kwargs.get('validate')

    jid = kwargs.get('jid', '*')

    scale = kwargs.get('scale')



    if fastforward: print(f"fast-forwarding {fastforward} seconds")
@@ -148,6 +150,8 @@ def load_data_from_df(jobs_df: pd.DataFrame, **kwargs): 
        else: # Prescribed replay

            scheduled_nodes = (jobs_df.loc[jidx, 'nodes']).tolist()

            

        if scale > 0: nodes_required = random.randint(1, scale)



        if gpu_trace.size > 0 and time_offset >= 0:

            job_info = job_dict(nodes_required, name, cpu_trace, gpu_trace, [], [], wall_time,

                                end_state, scheduled_nodes, time_offset, job_id, priority)