service-capacity-modeling 0.3.88__py3-none-any.whl → 0.3.90__py3-none-any.whl

service_capacity_modeling/enum_utils.py

@@ -5,6 +5,7 @@ runtime-accessible docstrings.
 
 import ast
 import inspect
+import sys
 from enum import Enum
 from functools import partial
 from operator import is_
@@ -16,6 +17,47 @@ from pydantic.json_schema import JsonSchemaValue
 from pydantic_core import CoreSchema
 
 
+__all__ = ["StrEnum", "enum_docstrings"]
+
+# StrEnum backport for Python 3.10 compatibility
+# On Python 3.11+, use the stdlib version
+if sys.version_info >= (3, 11):
+    from enum import StrEnum as StrEnum  # pylint: disable=useless-import-alias
+else:
+
+    class StrEnum(str, Enum):
+        """Backport of Python 3.11 StrEnum.
+
+        Provides consistent string behavior across all Python versions:
+        - f"{x}" returns the value (not "Foo.BAR")
+        - str(x) returns the value (not "Foo.BAR")
+        - x == "value" returns True (string comparison works)
+
+        This addresses PEP 663 which changed str(Enum) behavior in Python 3.11,
+        making (str, Enum) return "Foo.BAR" in f-strings instead of the value.
+        """
+
+        def __new__(cls, value: str, *args: Any, **kwargs: Any) -> "StrEnum":
+            if not isinstance(value, str):
+                raise TypeError(f"{value!r} is not a string")
+            member = str.__new__(cls, value)
+            member._value_ = value
+            return member
+
+        def __str__(self) -> str:
+            return str(self.value)
+
+        def __format__(self, format_spec: str) -> str:
+            # Ensures f-strings return value, not "Foo.BAR"
+            return str(self.value).__format__(format_spec)
+
+        @staticmethod
+        def _generate_next_value_(
+            name: str, start: int, count: int, last_values: list[str]
+        ) -> str:
+            return name.lower()
+
+
 E = TypeVar("E", bound=Enum)
 
 

service_capacity_modeling/interface.py

@@ -4,7 +4,6 @@ from __future__ import annotations
 import re
 import sys
 from decimal import Decimal
-from enum import Enum
 from fractions import Fraction
 from functools import lru_cache
 from typing import Any
@@ -22,6 +21,7 @@ from pydantic import ConfigDict
 from pydantic import Field
 
 from service_capacity_modeling.enum_utils import enum_docstrings
+from service_capacity_modeling.enum_utils import StrEnum
 
 GIB_IN_BYTES = 1024 * 1024 * 1024
 MIB_IN_BYTES = 1024 * 1024
@@ -46,16 +46,13 @@ class ExcludeUnsetModel(BaseModel):
 
 
 @enum_docstrings
-class IntervalModel(str, Enum):
+class IntervalModel(StrEnum):
     """Statistical distribution models for approximating intervals
 
     When we have uncertainty intervals (low, mid, high), we need to choose
     a probability distribution to model that uncertainty for simulation purposes.
     """
 
-    def __str__(self) -> str:
-        return str(self.value)
-
     def __repr__(self) -> str:
         return f"D({self.value})"
 
@@ -204,7 +201,7 @@ def normalized_aws_size(name: str) -> Fraction:
 ###############################################################################
 
 
-class Lifecycle(str, Enum):
+class Lifecycle(StrEnum):
     """Represents the lifecycle of hardware from initial preview
     to end-of-life.
 
@@ -223,7 +220,7 @@ class Lifecycle(str, Enum):
 
 
 @enum_docstrings
-class DriveType(str, Enum):
+class DriveType(StrEnum):
     """Represents the type and attachment model of storage drives
 
     Drives can be either local (ephemeral, instance-attached) or network-attached
@@ -366,7 +363,7 @@ class Drive(ExcludeUnsetModel):
 
 
 @enum_docstrings
-class Platform(str, Enum):
+class Platform(StrEnum):
     """Represents the CPU architecture or managed service platform
 
     Hardware can run on different CPU architectures (x86_64, ARM) or be a fully
@@ -617,7 +614,7 @@ class Pricing(ExcludeUnsetModel):
 
 
 @enum_docstrings
-class AccessPattern(str, Enum):
+class AccessPattern(StrEnum):
     """The access pattern determines capacity planning priorities: latency-sensitive
     services target low P99 latency, while throughput-oriented services optimize
     for maximum requests per second.
@@ -633,7 +630,7 @@ class AccessPattern(str, Enum):
 
 
 @enum_docstrings
-class AccessConsistency(str, Enum):
+class AccessConsistency(StrEnum):
     """
     Generally speaking consistency is expensive, so models need to know what
     kind of consistency will be required in order to estimate CPU usage
@@ -851,7 +848,7 @@ class CurrentClusters(ExcludeUnsetModel):
 
 
 @enum_docstrings
-class BufferComponent(str, Enum):
+class BufferComponent(StrEnum):
     """Represents well known buffer components such as compute and storage
 
     Note that while these are common and defined here for models to share,
@@ -887,7 +884,7 @@ class BufferComponent(str, Enum):
 
 
 @enum_docstrings
-class BufferIntent(str, Enum):
+class BufferIntent(StrEnum):
     """Defines the intent of buffer directives for capacity planning"""
 
     desired = "desired"

service_capacity_modeling/models/org/netflix/cassandra.py

@@ -638,8 +638,14 @@ def _target_rf(desires: CapacityDesires, user_copies: Optional[int]) -> int:
 
 
 class NflxCassandraArguments(BaseModel):
-    copies_per_region: int = Field(
-        default=3,
+    """Configuration arguments for the Netflix Cassandra capacity model.
+
+    This model centralizes all tunable parameters with their defaults.
+    Use `from_extra_model_arguments()` to parse a dict into a validated instance.
+    """
+
+    copies_per_region: Optional[int] = Field(
+        default=None,
         description="How many copies of the data will exist e.g. RF=3. If unsupplied"
         " this will be deduced from durability and consistency desires",
     )
@@ -663,9 +669,13 @@ class NflxCassandraArguments(BaseModel):
         default=192,
         description="What is the maximum size of a cluster in this region",
     )
-    max_local_disk_gib: int = Field(
-        default=5120,
-        description="The maximum amount of data we store per machine",
+    max_local_data_per_node_gib: int = Field(
+        default=1280,
+        description="Maximum data per node for local disk instances (GiB)",
+    )
+    max_attached_data_per_node_gib: int = Field(
+        default=2048,
+        description="Maximum data per node for attached disk instances (GiB)",
     )
     max_write_buffer_percent: float = Field(
         default=0.25,
@@ -680,6 +690,26 @@ class NflxCassandraArguments(BaseModel):
         "automatically adjust to 0.2",
     )
 
+    @classmethod
+    def from_extra_model_arguments(
+        cls, extra_model_arguments: Dict[str, Any]
+    ) -> "NflxCassandraArguments":
+        """Parse extra_model_arguments dict into a validated NflxCassandraArguments.
+
+        This centralizes default values - any field not in extra_model_arguments
+        will use the default defined in this model.
+
+        Handles legacy field name mappings:
+        - max_local_disk_gib -> max_local_data_per_node_gib (if not explicitly set)
+        """
+        # Handle legacy field name: max_local_disk_gib -> max_local_data_per_node_gib
+        args = dict(extra_model_arguments)
+        if "max_local_data_per_node_gib" not in args and "max_local_disk_gib" in args:
+            args["max_local_data_per_node_gib"] = args["max_local_disk_gib"]
+
+        # Pydantic will use defaults for any missing fields
+        return cls.model_validate(args)
+
 
 class NflxCassandraCapacityModel(CapacityModel):
     def __init__(self) -> None:
@@ -722,40 +752,22 @@ class NflxCassandraCapacityModel(CapacityModel):
         desires: CapacityDesires,
         extra_model_arguments: Dict[str, Any],
     ) -> Optional[CapacityPlan]:
-        # TODO: Standardize these extra model argument defaults in a single
-        # place. Many of them are defined here and as default values in the
-        # downstream method but only these ones are used which is confusing for
-        # readability
-
-        # Use durabiliy and consistency to compute RF.
-        copies_per_region = _target_rf(
-            desires, extra_model_arguments.get("copies_per_region", None)
-        )
-        require_local_disks: bool = extra_model_arguments.get(
-            "require_local_disks", False
-        )
-        require_attached_disks: bool = extra_model_arguments.get(
-            "require_attached_disks", False
-        )
+        # Parse extra_model_arguments into a validated model with centralized defaults
+        args = NflxCassandraArguments.from_extra_model_arguments(extra_model_arguments)
+
+        # Use durability and consistency to compute RF if not explicitly set
+        copies_per_region = _target_rf(desires, args.copies_per_region)
+
+        # Validate required_cluster_size for critical tiers
         required_cluster_size: Optional[int] = (
             NflxCassandraCapacityModel.get_required_cluster_size(
                 desires.service_tier, extra_model_arguments
             )
         )
 
-        max_rps_to_disk: int = extra_model_arguments.get("max_rps_to_disk", 500)
-        max_regional_size: int = extra_model_arguments.get("max_regional_size", 192)
-        max_local_data_per_node_gib: int = extra_model_arguments.get(
-            "max_local_data_per_node_gib",
-            extra_model_arguments.get("max_local_disk_gib", 1280),
-        )
-
-        max_write_buffer_percent: float = min(
-            0.5, extra_model_arguments.get("max_write_buffer_percent", 0.25)
-        )
-        max_table_buffer_percent: float = min(
-            0.5, extra_model_arguments.get("max_table_buffer_percent", 0.11)
-        )
+        # Apply caps to buffer percentages
+        max_write_buffer_percent = min(0.5, args.max_write_buffer_percent)
+        max_table_buffer_percent = min(0.5, args.max_table_buffer_percent)
 
         # Adjust heap defaults for high write clusters
         if (
@@ -772,12 +784,12 @@ class NflxCassandraCapacityModel(CapacityModel):
             desires=desires,
             zones_per_region=context.zones_in_region,
             copies_per_region=copies_per_region,
-            require_local_disks=require_local_disks,
-            require_attached_disks=require_attached_disks,
+            require_local_disks=args.require_local_disks,
+            require_attached_disks=args.require_attached_disks,
             required_cluster_size=required_cluster_size,
-            max_rps_to_disk=max_rps_to_disk,
-            max_regional_size=max_regional_size,
-            max_local_data_per_node_gib=max_local_data_per_node_gib,
+            max_rps_to_disk=args.max_rps_to_disk,
+            max_regional_size=args.max_regional_size,
+            max_local_data_per_node_gib=args.max_local_data_per_node_gib,
             max_write_buffer_percent=max_write_buffer_percent,
             max_table_buffer_percent=max_table_buffer_percent,
         )

service_capacity_modeling/models/org/netflix/counter.py

@@ -1,5 +1,4 @@
 from datetime import timedelta
-from enum import Enum
 from typing import Any
 from typing import Callable
 from typing import Dict
@@ -10,6 +9,7 @@ from pydantic import Field
 
 from .stateless_java import nflx_java_app_capacity_model
 from .stateless_java import NflxJavaAppArguments
+from service_capacity_modeling.enum_utils import StrEnum
 from service_capacity_modeling.interface import AccessConsistency
 from service_capacity_modeling.interface import AccessPattern
 from service_capacity_modeling.interface import CapacityDesires
@@ -26,13 +26,13 @@ from service_capacity_modeling.interface import RegionContext
 from service_capacity_modeling.models import CapacityModel
 
 
-class NflxCounterCardinality(Enum):
+class NflxCounterCardinality(StrEnum):
     low = "low"
     medium = "medium"
     high = "high"
 
 
-class NflxCounterMode(Enum):
+class NflxCounterMode(StrEnum):
     best_effort = "best-effort"
     eventual = "eventual"
     exact = "exact"
@@ -95,7 +95,7 @@ class NflxCounterCapacityModel(CapacityModel):
     ) -> Tuple[Tuple[str, Callable[[CapacityDesires], CapacityDesires]], ...]:
         stores = []
 
-        if extra_model_arguments["counter.mode"] == NflxCounterMode.best_effort.value:
+        if extra_model_arguments["counter.mode"] == NflxCounterMode.best_effort:
             stores.append(("org.netflix.evcache", lambda x: x))
         else:
             # Shared evcache cluster is used for eventual and exact counters
@@ -114,9 +114,9 @@ class NflxCounterCapacityModel(CapacityModel):
                 # high cardinality : rollups happen once every 10 seconds
                 # TODO: Account for read amplification from time slice configs
                 #       for better model accuracy
-                if counter_cardinality == NflxCounterCardinality.low.value:
+                if counter_cardinality == NflxCounterCardinality.low:
                     rollups_per_second = counter_deltas_per_second.scale(0.0167)
-                elif counter_cardinality == NflxCounterCardinality.medium.value:
+                elif counter_cardinality == NflxCounterCardinality.medium:
                     rollups_per_second = counter_deltas_per_second.scale(0.0333)
                 else:
                     rollups_per_second = counter_deltas_per_second.scale(0.1)

service_capacity_modeling/models/org/netflix/evcache.py

@@ -1,6 +1,5 @@
 import logging
 import math
-from enum import Enum
 from typing import Any
 from typing import Dict
 from typing import Optional
@@ -9,6 +8,7 @@ from typing import Tuple
 from pydantic import BaseModel
 from pydantic import Field
 
+from service_capacity_modeling.enum_utils import StrEnum
 from service_capacity_modeling.interface import AccessConsistency
 from service_capacity_modeling.interface import AccessPattern
 from service_capacity_modeling.interface import Buffer
@@ -45,7 +45,7 @@ from service_capacity_modeling.stats import dist_for_interval
 logger = logging.getLogger(__name__)
 
 
-class Replication(str, Enum):
+class Replication(StrEnum):
     none = "none"
     sets = "sets"
     evicts = "evicts"

service_capacity_modeling/models/org/netflix/kafka.py

@@ -1,6 +1,5 @@
 import logging
 import math
-from enum import Enum
 from typing import Any
 from typing import Dict
 from typing import Optional
@@ -9,6 +8,7 @@ from typing import Tuple
 from pydantic import BaseModel
 from pydantic import Field
 
+from service_capacity_modeling.enum_utils import StrEnum
 from service_capacity_modeling.interface import AccessConsistency
 from service_capacity_modeling.interface import AccessPattern
 from service_capacity_modeling.interface import Buffer
@@ -45,7 +45,7 @@ from service_capacity_modeling.models.org.netflix.iso_date_math import iso_to_se
 logger = logging.getLogger(__name__)
 
 
-class ClusterType(str, Enum):
+class ClusterType(StrEnum):
     strong = "strong"
     ha = "high-availability"
 

{service_capacity_modeling-0.3.88.dist-info → service_capacity_modeling-0.3.90.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: service-capacity-modeling
-Version: 0.3.88
+Version: 0.3.90
 Summary: Contains utilities for modeling capacity for pluggable workloads
 Author: Joseph Lynch
 Author-email: josephl@netflix.com

{service_capacity_modeling-0.3.88.dist-info → service_capacity_modeling-0.3.90.dist-info}/RECORD

@@ -1,7 +1,7 @@
 service_capacity_modeling/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 service_capacity_modeling/capacity_planner.py,sha256=JRagEFlg3u_zB1N5GzGsKAN55JZLad6p4IF_PmL8kcg,32780
-service_capacity_modeling/enum_utils.py,sha256=-jf9IkSg28B8ekXXel1l34FZw5iF525EttyiVft0MLg,3768
-service_capacity_modeling/interface.py,sha256=AV7QlogbCrdjl4nMQQhouyy-2_jNmngU6081NxEC5o4,42370
+service_capacity_modeling/enum_utils.py,sha256=50Rw2kgYoJYCrybSbo9WaPPCWxlF5CyPCQtHxQ3kB18,5229
+service_capacity_modeling/interface.py,sha256=mgX2qUsYf_rVxbYg1rFjRUH-QY11Z87mke4nWa1LwhM,42327
 service_capacity_modeling/stats.py,sha256=LCNUcQPfwF5hhIZwsfAsDe4ZbnuhDnl3vQHKfpK61Xc,6142
 service_capacity_modeling/hardware/__init__.py,sha256=P5ostvoSOMUqPODtepeFYb4qfTVH0E73mMFraP49rYU,9196
 service_capacity_modeling/hardware/profiles/__init__.py,sha256=7-y3JbCBkgzaAjFla2RIymREcImdZ51HTl3yn3vzoGw,1602
@@ -53,17 +53,17 @@ service_capacity_modeling/models/utils.py,sha256=WosEEg4o1_WSbTb5mL-M1v8JuWJgvS2
 service_capacity_modeling/models/org/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 service_capacity_modeling/models/org/netflix/__init__.py,sha256=2Ld2NPxiO3vbYtOMqHtrV4f1nEH390Hoxqo3I5NbBDI,2553
 service_capacity_modeling/models/org/netflix/aurora.py,sha256=Js33ZjxCtt34HiDPsWRT9mjKCAsnnCo9du15QArVFMo,13073
-service_capacity_modeling/models/org/netflix/cassandra.py,sha256=Fp37bHtWRJctJYYJhW78YbLeFXoX26QxREP_BMEItNE,39003
+service_capacity_modeling/models/org/netflix/cassandra.py,sha256=MUILIhdxlqIxOwMbFJvy7luRYqBxraIkIVU_ZKnjxOo,39627
 service_capacity_modeling/models/org/netflix/control.py,sha256=wkJnqG7Nn7kQNcKEMzl9LFlA9xTAji2wYxL011VAKPI,5762
-service_capacity_modeling/models/org/netflix/counter.py,sha256=XopKlNMdvO5EbxAggn-KW_q7L3aKtXLXbry4ocl6i5Q,10494
+service_capacity_modeling/models/org/netflix/counter.py,sha256=kTDL7dCnkn-XU27_Z1VBc4CCLCPoOqJZe9WgcENHHd4,10517
 service_capacity_modeling/models/org/netflix/crdb.py,sha256=iW7tyG8jpXhHIdXrw3DPYSHRAknPN42MlCRLJO4o9C8,20826
 service_capacity_modeling/models/org/netflix/ddb.py,sha256=9qRiuTqWev9zbYFFzewyowU7M41uALsuLklYx20yAXw,26502
 service_capacity_modeling/models/org/netflix/elasticsearch.py,sha256=zPrC6b2LNrAh3IWE3HCMUEYASacjYbHChbO4WZSMma4,25234
 service_capacity_modeling/models/org/netflix/entity.py,sha256=VHgEwnGtJAKlhvbE2kTif75OZmIsjjjoZrT6kb1LTgA,8750
-service_capacity_modeling/models/org/netflix/evcache.py,sha256=x8KwoULVpDm8UXLXEXfH8_zvoBDs8jSfOQLkaWKFcOg,25654
+service_capacity_modeling/models/org/netflix/evcache.py,sha256=BDVRWely3F_3Ecb3Um3dQ024_I6XgvagpRJ6zdP5E18,25687
 service_capacity_modeling/models/org/netflix/graphkv.py,sha256=7ncEhx9lLsN_vGIKNHkvWfDdKffG7cYe91Wr-DB7IjU,8659
 service_capacity_modeling/models/org/netflix/iso_date_math.py,sha256=oC5sgIXDqwOp6-5z2bdTkm-bJLlnzhqcONI_tspHjac,1137
-service_capacity_modeling/models/org/netflix/kafka.py,sha256=JwiRsXOK5hYEr2YZL0SpHK2Z61PhbrItHXitAPdzty8,25514
+service_capacity_modeling/models/org/netflix/kafka.py,sha256=1Ut1yBHgMCdH5chByuqBSmrZF-XpKdvFxAhRHaRV9nQ,25547
 service_capacity_modeling/models/org/netflix/key_value.py,sha256=WH8NblHqHwnAAumB2Zz1Qd4NBFWDQEQ1rpBcP3fVVQk,9409
 service_capacity_modeling/models/org/netflix/postgres.py,sha256=LBxDqkc-lYxDBu2VwNLuf2Q4o4hU3jPwu4YSt33Oe-8,4128
 service_capacity_modeling/models/org/netflix/rds.py,sha256=8GVmpMhTisZPdT-mP1Sx5U7VAF32lnTI27iYPfGg9CY,10930
@@ -77,9 +77,9 @@ service_capacity_modeling/tools/auto_shape.py,sha256=Jx9H2ay9-H_kUDjtB141owQNxGF
 service_capacity_modeling/tools/fetch_pricing.py,sha256=Qp-XMymkY1dvtyS51RufmEpfgOHv-IQ-XyzS8wp2-qM,4021
 service_capacity_modeling/tools/generate_missing.py,sha256=F7YqvMJAV4nZc20GNrlIsnQSF8_77sLgwYZqc5k4LDg,3099
 service_capacity_modeling/tools/instance_families.py,sha256=e5RuYkCLUITvsAazDH12B6KjX_PaBsv6Ne3mj0HK_sQ,9223
-service_capacity_modeling-0.3.88.dist-info/licenses/LICENSE,sha256=nl_Lt5v9VvJ-5lWJDT4ddKAG-VZ-2IaLmbzpgYDz2hU,11343
-service_capacity_modeling-0.3.88.dist-info/METADATA,sha256=tJJaylvAZ5SfJqXD8F_Sz_HgEO2-Q0Z9HBuUmPKguVU,10366
-service_capacity_modeling-0.3.88.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
-service_capacity_modeling-0.3.88.dist-info/entry_points.txt,sha256=ZsjzpG5SomWpT1zCE19n1uSXKH2gTI_yc33sdl0vmJg,146
-service_capacity_modeling-0.3.88.dist-info/top_level.txt,sha256=H8XjTCLgR3enHq5t3bIbxt9SeUkUT8HT_SDv2dgIT_A,26
-service_capacity_modeling-0.3.88.dist-info/RECORD,,
+service_capacity_modeling-0.3.90.dist-info/licenses/LICENSE,sha256=nl_Lt5v9VvJ-5lWJDT4ddKAG-VZ-2IaLmbzpgYDz2hU,11343
+service_capacity_modeling-0.3.90.dist-info/METADATA,sha256=QAmQcpV4CC-6rKvYKnmzrb4OjpW7sCQdaFAdB5u9TVo,10366
+service_capacity_modeling-0.3.90.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+service_capacity_modeling-0.3.90.dist-info/entry_points.txt,sha256=ZsjzpG5SomWpT1zCE19n1uSXKH2gTI_yc33sdl0vmJg,146
+service_capacity_modeling-0.3.90.dist-info/top_level.txt,sha256=H8XjTCLgR3enHq5t3bIbxt9SeUkUT8HT_SDv2dgIT_A,26
+service_capacity_modeling-0.3.90.dist-info/RECORD,,