This page intentionally left blank.

dummy slide

Harnessing Spare Cores to
Breeze Through Cloud Compute

Gergely Daróczi

Spare Cores Team

Slides: sparecores.com/talks

Press Space or click the green arrow icons to navigate the slides ->

>>> from sparecores import why

Data Science / Machine Learning batch jobs:

  • run SQL
  • run R or Python script
    • train a simple model, reporting, API integrations etc.
    • train hierarchical models/GBMs/neural nets etc.

>>> from sparecores import why

Data Science / Machine Learning batch jobs:

  • run SQL
  • run R or Python script
    • train a simple model, reporting, API integrations etc.
    • train hierarchical models/GBMs/neural nets etc.

Scaling (DS) infrastructure.

>>> from sparecores import why

>>> from sparecores import why

>>> from sparecores import why

AWS ECS

AWS Batch

Kubernetes

>>> from sparecores import why

Source: xkcd

>>> from sparecores import why

>>> from sparecores import why

>>> from sparecores import why

>>> from sparecores import why

Other use-cases:

  • stats/ML/AI model training,
  • ETL pipelines,
  • traditional CI/CD workflows for compiling and testing software,
  • building Docker images,
  • rendering images and videos,
  • etc.

>>> from sparecores import why

>>> from sparecores import intro

  • Open-source tools, database schemas and documentation to inspect and inventory cloud vendors and their compute resource offerings.
  • Managed infrastructure, databases, APIs, SDKs, and web applications to make these data sources publicly accessible.
  • Helpers to start and manage instances in your own environment.
  • SaaS to run containers in a managed environment without direct vendor engagement.

>>> from sparecores import intro

Source: sparecores.com

>>> from sparecores import intro

>>> from sparecores import intro

>>> from sparecores import intro

>>> from sparecores import intro

>>> from sparecores import intro

>>> from sparecores import intro

>>> from sparecores import intro

>>> from sparecores import intro

>>> from rich import print as pp
>>> from sc_crawler.tables import Server
>>> from sqlmodel import create_engine, Session, select
>>> engine = create_engine("sqlite:///sc-data-all.db")
>>> session = Session(engine)
>>> server = session.exec(select(Server).where(Server.server_id == 'g4dn.xlarge')).one()
>>> pp(server)
Server(
    server_id='g4dn.xlarge',
    vendor_id='aws',
    display_name='g4dn.xlarge',
    api_reference='g4dn.xlarge',
    name='g4dn.xlarge',
    family='g4dn',
    description='Graphics intensive [Instance store volumes] [Network and EBS optimized] Gen4 xlarge',

    status=<Status.ACTIVE: 'active'>,
    observed_at=datetime.datetime(2024, 6, 6, 10, 18, 4, 127254),

    hypervisor='nitro',
    vcpus=4,
    cpu_cores=2,
    cpu_allocation=<CpuAllocation.DEDICATED: 'Dedicated'>,
    cpu_manufacturer='Intel',
    cpu_family='Xeon',
    cpu_model='8259CL',
    cpu_architecture=<CpuArchitecture.X86_64: 'x86_64'>,
    cpu_speed=3.5,
    cpu_l1_cache=None,
    cpu_l2_cache=None,
    cpu_l3_cache=None,
    cpu_flags=[],

    memory_amount=16384,
    memory_generation=<DdrGeneration.DDR4: 'DDR4'>,
    memory_speed=3200,
    memory_ecc=None,

    gpu_count=1,
    gpu_memory_min=16384,
    gpu_memory_total=16384,
    gpu_manufacturer='Nvidia',
    gpu_family='Turing',
    gpu_model='Tesla T4',
    gpus=[
        {
            'manufacturer': 'Nvidia',
            'family': 'Turing',
            'model': 'Tesla T4',
            'memory': 15360,
            'firmware_version': '535.171.04',
            'bios_version': '90.04.96.00.A0',
            'graphics_clock': 1590,
            'sm_clock': 1590,
            'mem_clock': 5001,
            'video_clock': 1470
        }
    ],

    storage_size=125,
    storage_type=<StorageType.NVME_SSD: 'nvme ssd'>,
    storages=[{'size': 125, 'storage_type': 'nvme ssd'}],

    network_speed=5.0,
    inbound_traffic=0.0,
    outbound_traffic=0.0,
    ipv4=0,
)

>>> sparecores.__dir__()

>>> import sc_crawler

  • ETL framework with database schema and inventory method definitions
  • Database migration tool supporting multiple database engines
  • Manual list of vendors and metadata
  • Vendor API integrations to list regions, zones, servers, storages, prices, included free traffic and IPv4 addresses etc.
  • Spare Cores Inspector integration for hardware discovery and benchmark scores
  • Dependency for other Spare Cores components (schemas)

>>> import sc_crawler

>>> from sc_crawler import fks

Source: dbdocs.io/spare-cores

>>> from sc_crawler import scd

Need to optionally track price etc. changes.

class Scd(ScModel):
    """Override the `observed_at` column to be primary key in SCD tables."""

    observed_at: datetime = Field(
        primary_key=True,
        default_factory=datetime.utcnow,
        sa_column_kwargs={"onupdate": datetime.utcnow},
        description="Timestamp of the last observation.",
    )

>>> from sc_crawler import alembic

Let’s update the cpu_cores column to be optional, as some vendors as shy sharing that over their APIs. We will backfill with the Spare Cores Inspector!

"""v0.1.1 cores optional

Revision ID: 4691089690c2
Revises: 98894dffd37c
Create Date: 2024-04-10 00:59:03.509522

"""

from typing import Sequence, Union

import sqlalchemy as sa
import sqlmodel
from alembic import op

# revision identifiers, used by Alembic.
revision: str = "4691089690c2"
down_revision: Union[str, None] = "98894dffd37c"
branch_labels: Union[str, Sequence[str], None] = None
depends_on: Union[str, Sequence[str], None] = None

## need to provide the table schema for offline mode support
meta = sa.MetaData()
server_table = sa.Table(
    "server_scd" if op.get_context().config.attributes.get("scd") else "server",
    meta,
    sa.Column(
        "vendor_id",
        sqlmodel.sql.sqltypes.AutoString(),
        nullable=False,
    ),
    sa.Column(
        "server_id",
        sqlmodel.sql.sqltypes.AutoString(),
        nullable=False,
    ),
    sa.Column(
        "name",
        sqlmodel.sql.sqltypes.AutoString(),
        nullable=False,
    ),
    sa.Column(
        "vcpus",
        sa.Integer(),
        nullable=False,
    ),
    sa.Column(
        "hypervisor",
        sqlmodel.sql.sqltypes.AutoString(),
        nullable=True,
    ),
    sa.Column(
        "cpu_allocation",
        sa.Enum("SHARED", "BURSTABLE", "DEDICATED", name="cpuallocation"),
        nullable=False,
    ),
    sa.Column(
        "cpu_cores",
        sa.Integer(),
        nullable=False,
    ),
    sa.Column(
        "cpu_speed",
        sa.Float(),
        nullable=True,
    ),
    sa.Column(
        "cpu_architecture",
        sa.Enum(
            "ARM64",
            "ARM64_MAC",
            "I386",
            "X86_64",
            "X86_64_MAC",
            name="cpuarchitecture",
        ),
        nullable=False,
    ),
    sa.Column(
        "cpu_manufacturer",
        sqlmodel.sql.sqltypes.AutoString(),
        nullable=True,
    ),
    sa.Column(
        "cpu_family",
        sqlmodel.sql.sqltypes.AutoString(),
        nullable=True,
    ),
    sa.Column(
        "cpu_model",
        sqlmodel.sql.sqltypes.AutoString(),
        nullable=True,
    ),
    sa.Column(
        "cpus",
        sa.JSON(),
        nullable=False,
    ),
    sa.Column("memory", sa.Integer(), nullable=False),
    sa.Column(
        "gpu_count",
        sa.Integer(),
        nullable=False,
    ),
    sa.Column(
        "gpu_memory_min",
        sa.Integer(),
        nullable=True,
    ),
    sa.Column(
        "gpu_memory_total",
        sa.Integer(),
        nullable=True,
    ),
    sa.Column(
        "gpu_manufacturer",
        sqlmodel.sql.sqltypes.AutoString(),
        nullable=True,
    ),
    sa.Column(
        "gpu_model",
        sqlmodel.sql.sqltypes.AutoString(),
        nullable=True,
    ),
    sa.Column(
        "gpus",
        sa.JSON(),
        nullable=False,
    ),
    sa.Column(
        "storage_size",
        sa.Integer(),
        nullable=False,
    ),
    sa.Column(
        "storage_type",
        sa.Enum("HDD", "SSD", "NVME_SSD", "NETWORK", name="storagetype"),
        nullable=True,
    ),
    sa.Column(
        "storages",
        sa.JSON(),
        nullable=False,
    ),
    sa.Column(
        "network_speed",
        sa.Float(),
        nullable=True,
    ),
    sa.Column(
        "inbound_traffic",
        sa.Float(),
        nullable=False,
    ),
    sa.Column(
        "outbound_traffic",
        sa.Float(),
        nullable=False,
    ),
    sa.Column(
        "ipv4",
        sa.Integer(),
        nullable=False,
    ),
    sa.Column(
        "status",
        sa.Enum("ACTIVE", "INACTIVE", name="status"),
        nullable=False,
    ),
    sa.Column(
        "observed_at",
        sa.DateTime(),
        nullable=False,
    ),
    sa.ForeignKeyConstraint(
        ["vendor_id"],
        ["vendor.vendor_id"],
    ),
    sa.PrimaryKeyConstraint("vendor_id", "server_id", "observed_at")
    if op.get_context().config.attributes.get("scd")
    else sa.PrimaryKeyConstraint("vendor_id", "server_id"),
)

def upgrade() -> None:
    if op.get_context().config.attributes.get("scd"):
        with op.batch_alter_table(
            "server_scd", schema=None, copy_from=server_table
        ) as batch_op:
            batch_op.alter_column(
                "cpu_cores", existing_type=sa.INTEGER(), nullable=True
            )
    else:
        with op.batch_alter_table(
            "server", schema=None, copy_from=server_table
        ) as batch_op:
            batch_op.alter_column(
                "cpu_cores", existing_type=sa.INTEGER(), nullable=True
            )


def downgrade() -> None:
    if op.get_context().config.attributes.get("scd"):
        with op.batch_alter_table(
            "server_scd", schema=None, copy_from=server_table
        ) as batch_op:
            batch_op.alter_column(
                "cpu_cores", existing_type=sa.INTEGER(), nullable=False
            )
    else:
        with op.batch_alter_table(
            "server", schema=None, copy_from=server_table
        ) as batch_op:
            batch_op.alter_column(
                "cpu_cores", existing_type=sa.INTEGER(), nullable=False
            )

>>> from sc_crawler import alembic

$ sc-crawler schemas upgrade --sql
CREATE TABLE _alembic_tmp_server (
    vendor_id VARCHAR NOT NULL,
    server_id VARCHAR NOT NULL,
    name VARCHAR NOT NULL,
    vcpus INTEGER NOT NULL,
    hypervisor VARCHAR,
    cpu_allocation VARCHAR(9) NOT NULL,
    cpu_cores INTEGER,
    cpu_speed FLOAT,
    cpu_architecture VARCHAR(10) NOT NULL,
    cpu_manufacturer VARCHAR,
    cpu_family VARCHAR,
    cpu_model VARCHAR,
    cpus JSON NOT NULL,
    memory INTEGER NOT NULL,
    gpu_count INTEGER NOT NULL,
    gpu_memory_min INTEGER,
    gpu_memory_total INTEGER,
    gpu_manufacturer VARCHAR,
    gpu_model VARCHAR,
    gpus JSON NOT NULL,
    storage_size INTEGER NOT NULL,
    storage_type VARCHAR(8),
    storages JSON NOT NULL,
    network_speed FLOAT,
    inbound_traffic FLOAT NOT NULL,
    outbound_traffic FLOAT NOT NULL,
    ipv4 INTEGER NOT NULL,
    status VARCHAR(8) NOT NULL,
    observed_at DATETIME NOT NULL,
    description VARCHAR,
    PRIMARY KEY (vendor_id, server_id),
    FOREIGN KEY(vendor_id) REFERENCES vendor (vendor_id)
);

INSERT INTO _alembic_tmp_server (vendor_id, server_id, name, vcpus, hypervisor, cpu_allocation, cpu_cores, cpu_speed, cpu_architecture, cpu_manufacturer, cpu_family, cpu_model, cpus, memory, gpu_count, gpu_memory_min, gpu_memory_total, gpu
_manufacturer, gpu_model, gpus, storage_size, storage_type, storages, network_speed, inbound_traffic, outbound_traffic, ipv4, status, observed_at) SELECT server.vendor_id, server.server_id, server.name, server.vcpus, server.hypervisor, ser
ver.cpu_allocation, server.cpu_cores, server.cpu_speed, server.cpu_architecture, server.cpu_manufacturer, server.cpu_family, server.cpu_model, server.cpus, server.memory, server.gpu_count, server.gpu_memory_min, server.gpu_memory_total, se
rver.gpu_manufacturer, server.gpu_model, server.gpus, server.storage_size, server.storage_type, server.storages, server.network_speed, server.inbound_traffic, server.outbound_traffic, server.ipv4, server.status, server.observed_at
FROM server;

DROP TABLE server;

ALTER TABLE _alembic_tmp_server RENAME TO server;

UPDATE alembic_version SET version_num='4691089690c2' WHERE alembic_version.version_num = '98894dffd37c';

>>> from sc_crawler import hwinfo

  • Varying quality and availability of data at different vendors.

>>> from sc_crawler import hwinfo

  • Varying quality and availability of data at different vendors.

>>> from sc_crawler import hwinfo

  • Varying quality and availability of data at different vendors.
    • No SSD info via the API. Parse from server description! 🙀
    • No CPU info via the API. Extracting from homepage! 😿
    • No hypervisor info via the API. Manual mappings! 🙀😿
  • Region, right?
    • ID, eg eu-west-1
    • Name, eg Europe (Ireland)
    • Alias, eg EU (Ireland)
    • API reference, eg eu-west-1
    • Display name, eg Dublin (IE)
    • Exact location? Energy source?

>>> from sc_crawler import pricing

  • No way to find SKUs by filtering in the API call. Get all, search locally.

  • f1-micro is one out of 2 instances with simple pricing.

    • For other instances, lookup SKUs for CPU + RAM and do the math.
  • Match instance family with SKU via search in description, e.g. C2D.
  • Except for c2, which is called “Compute optimized”.
  • And m2 is actually priced at a premium on the top of m1.
  • The n1 resource group is not CPU/RAM, but N1Standard, extract if it’s CPU or RAM price from description.

>>> import sc_crawler

Source: dbhub.io/sparecores

>>> import sc_data

  • GitHub Action set up to run the Crawler every 5 minutes.
    • ~9000 GHA runs
    • ~100 releases (with non-price changes)
  • Make the data available in a public (CC BY-SA) SQLite database:
    • 60MiB SQLite
    • 1,120 active servers and their ~135k prices tracked
    • ~130k measured scores across 24 benchmarks
  • Thin Python package to keep the data updated from S3.
    • Package version is tied to Crawler version.

>>> import sc_inspector

Information collected from vendor APIs is very limited, so we run:

  • Hardware inspection tools:
    • dmidecode
    • lscpu
    • lshw
  • Benchmarking workloads:
    • bw_mem
    • Compression algos
    • OpenSSL hash functions and block ciphers
    • Geekbench 6
    • stress-ng

>>> import sc_inspector

Data is collected in public: sc-inspector-data repo on GitHub.

>>> import sc_inspector

>>> import sc_inspector

docker run --rm -ti -v /var/run/docker.sock:/var/run/docker.sock \
    -e GITHUB_TOKEN=${GITHUB_TOKEN} \
    -e BENCHMARK_SECRETS_PASSPHRASE=${BENCHMARK_SECRETS_PASSPHRASE} \
    ghcr.io/sparecores/sc-inspector:main \
    inspect --vendor ${VENDOR} --instance ${INSTANCE} --gpu-count ${GPU_COUNT}

>>> import sc_runner 

$ docker run --rm -ti \
  ghcr.io/sparecores/sc-runner:main \
    create aws --instance t4g.nano

Updating (aws.us-west-2.None.t4g.nano):

    pulumi:pulumi:Stack runner-aws.us-west-2.None.t4g.nano running
 +  pulumi:providers:aws us-west-2 creating (0s)
@ updating....
 +  pulumi:providers:aws us-west-2 created (0.29s)
 +  aws:ec2:SecurityGroup t4g.nano creating (0s)
@ updating.....
 +  aws:ec2:SecurityGroup t4g.nano created (2s)
@ updating....
 +  aws:vpc:SecurityGroupIngressRule t4g.nano-0 creating (0s)
 +  aws:vpc:SecurityGroupIngressRule t4g.nano-1 creating (0s)
 +  aws:ec2:Instance t4g.nano creating (0s)
 +  aws:vpc:SecurityGroupEgressRule t4g.nano-1 creating (0s)
 +  aws:vpc:SecurityGroupEgressRule t4g.nano-0 creating (0s)
@ updating....
 +  aws:vpc:SecurityGroupIngressRule t4g.nano-0 created (1s)
 +  aws:vpc:SecurityGroupIngressRule t4g.nano-1 created (1s)
 +  aws:vpc:SecurityGroupEgressRule t4g.nano-1 created (1s)
@ updating....
 +  aws:vpc:SecurityGroupEgressRule t4g.nano-0 created (1s)
@ updating..............
 +  aws:ec2:Instance t4g.nano created (13s)
@ updating....
    pulumi:pulumi:Stack runner-aws.us-west-2.None.t4g.nano
Resources:
    + 7 created
    1 unchanged

>>> import sc_runner

>>> import sc_runner

>>> import sc_keeper 

$ curl https://keeper.sparecores.net/server/aws/g4dn.xlarge | jq
  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100  111k  100  111k    0     0  79795      0  0:00:01  0:00:01 --:--:-- 79799
{
  "vendor_id": "aws",
  "server_id": "g4dn.xlarge",
  "name": "g4dn.xlarge",
  "api_reference": "g4dn.xlarge",
  "display_name": "g4dn.xlarge",
  "description": "Graphics intensive [Instance store volumes] [Network and EBS optimized] Gen4 xlarge",
  "family": "g4dn",
  "vcpus": 4,
  "hypervisor": "nitro",
  "cpu_allocation": "Dedicated",
  "cpu_cores": 2,
  "cpu_speed": 3.5,
  "cpu_architecture": "x86_64",
  "cpu_manufacturer": "Intel",
  "cpu_family": "Xeon",
  "cpu_model": "8259CL",
  "cpu_l1_cache": 131072,
  "cpu_l2_cache": 2097152,
  "cpu_l3_cache": 37486592,
  "cpu_flags": [
    "fpu",
    "vme",
    "de",
    "pse",
    "tsc",
    "msr",
    "pae",
    ...
  ],
  "cpus": [],
  "memory_amount": 16384,
  "memory_generation": "DDR4",
  "memory_speed": 2933,
  "memory_ecc": null,
  "gpu_count": 1,
  "gpu_memory_min": 16384,
  "gpu_memory_total": 16384,
  "gpu_manufacturer": "Nvidia",
  "gpu_family": "Turing",
  "gpu_model": "Tesla T4",
  "gpus": [
    {
      "manufacturer": "Nvidia",
      "family": "Turing",
      "model": "Tesla T4",
      "memory": 15360,
      "firmware_version": "535.171.04",
      "bios_version": "90.04.96.00.A0",
      "graphics_clock": 1590,
      "sm_clock": 1590,
      "mem_clock": 5001,
      "video_clock": 1470
    }
  ],
  "storage_size": 125,
  "storage_type": "nvme ssd",
  "storages": [
    {
      "size": 125,
      "storage_type": "nvme ssd"
    }
  ],
  "network_speed": 5.0,
  "inbound_traffic": 0.0,
  "outbound_traffic": 0.0,
  "ipv4": 0,
  "status": "active",
  "observed_at": "2024-06-09T21:20:22.005194",
  "vendor": {
    "logo": "https://sc-data-public-40e9d310.s3.amazonaws.com/cdn/logos/aws.svg",
    "address_line": "410 Terry Ave N",
    "name": "Amazon Web Services",
    "zip_code": "98109",
    "founding_year": 2002,
    "state": "Washington",
    "status_page": "https://health.aws.amazon.com/health/status",
    "vendor_id": "aws",
    "homepage": "https://aws.amazon.com",
    "country_id": "US",
    "status": "active",
    "observed_at": "2024-06-09T21:50:32.658281",
    "city": "Seattle"
  },
  "prices": [
    {
      "vendor_id": "aws",
      "region_id": "af-south-1",
      "zone_id": "afs1-az1",
      "server_id": "g4dn.xlarge",
      "operating_system": "Linux",
      "allocation": "ondemand",
      "unit": "hour",
      "price": 0.698,
      "price_upfront": 0.0,
      "price_tiered": [],
      "currency": "USD",
      "status": "active",
      "observed_at": "2024-06-09T21:21:10.015921",
      "region": {
        "country_id": "ZA",
        "state": null,
        "founding_year": 2020,
        "green_energy": false,
        "name": "Africa (Cape Town)",
        "city": "Cape Town",
        "status": "active",
        "address_line": null,
        "observed_at": "2024-06-09T21:19:37.529944",
        "zip_code": null,
        "lon": 18.3758801,
        "region_id": "af-south-1",
        "display_name": "Cape Town (ZA)",
        "lat": -33.914651,
        "vendor_id": "aws",
        "api_reference": "af-south-1",
        "aliases": []
      },
      "zone": {
        "region_id": "af-south-1",
        "zone_id": "afs1-az1",
        "api_reference": "af-south-1a",
        "status": "active",
        "vendor_id": "aws",
        "name": "af-south-1a",
        "display_name": "af-south-1a",
        "observed_at": "2024-06-09T21:19:40.425499"
      }
    },
    {
      "vendor_id": "aws",
      "region_id": "af-south-1",
      "zone_id": "afs1-az2",
      "server_id": "g4dn.xlarge",
      "operating_system": "Linux",
      "allocation": "spot",
      "unit": "hour",
      "price": 0.2251,
      "price_upfront": 0.0,
      "price_tiered": [],
      "currency": "USD",
      "status": "active",
      "observed_at": "2024-06-09T18:16:26",
      "region": {
        "country_id": "ZA",
        "state": null,
        "founding_year": 2020,
        "green_energy": false,
        "name": "Africa (Cape Town)",
        "city": "Cape Town",
        "status": "active",
        "address_line": null,
        "observed_at": "2024-06-09T21:19:37.529944",
        "zip_code": null,
        "lon": 18.3758801,
        "region_id": "af-south-1",
        "display_name": "Cape Town (ZA)",
        "lat": -33.914651,
        "vendor_id": "aws",
        "api_reference": "af-south-1",
        "aliases": []
      },
      "zone": {
        "region_id": "af-south-1",
        "zone_id": "afs1-az2",
        "api_reference": "af-south-1b",
        "status": "active",
        "vendor_id": "aws",
        "name": "af-south-1b",
        "display_name": "af-south-1b",
        "observed_at": "2024-06-09T21:19:40.425554"
      }
    },
    ...
  ],
  "benchmark_scores": [
    {
      "server_id": "g4dn.xlarge",
      "config": {},
      "status": "active",
      "vendor_id": "aws",
      "benchmark_id": "bogomips",
      "score": 5000.0,
      "note": null,
      "observed_at": "2024-06-07T04:26:48.643640"
    },
    {
      "server_id": "g4dn.xlarge",
      "config": {
        "cores": 1,
        "framework_version": "0.17.08"
      },
      "status": "active",
      "vendor_id": "aws",
      "benchmark_id": "stress_ng:cpu_all",
      "score": 1385.583093,
      "note": null,
      "observed_at": "2024-06-07T04:27:14.552982"
    },
    {
      "server_id": "g4dn.xlarge",
      "config": {
        "cores": 4,
        "framework_version": "0.17.08"
      },
      "status": "active",
      "vendor_id": "aws",
      "benchmark_id": "stress_ng:cpu_all",
      "score": 4013.022928,
      "note": null,
      "observed_at": "2024-06-07T04:27:02.508145"
    }
  ]
}

>>> import sc_keeper

>>> import sc_keeper

import { AppModule } from ‘sc-www’;

import { AppModule } from ‘sc-www’;

Source: sparecores.com

>>> import __future__

  • Add support for more vendors
    • Crawler (vendor API integration)
    • Runner (pulumi)
  • More SDKs (PyPI, npm, CRAN, etc.)
  • Data analysis, blog posts
  • My Spare Cores (dashboard)
  • SaaS 👀

>>> from sparecores import team

@bra-fsn

@palabola

@daroczig

>>> from sparecores import team

@bra-fsn

Infrastructure and Python veteran.

@palabola

Guardian of the front-end and Node.js tools.

@daroczig

Hack of all trades, master of NaN.

>>> from sparecores import support

Slides: sparecores.com/talks