Workload characterization-based capacity planning for cost-effective and high-performance serverless execution environment

A system for workload characterization-based capacity planning for a serverless execution environment, the system comprising:

• a memory storing instructions;
• one or more Input/Output (I/O) interfaces;
• and
• one or more hardware processors coupled to the memory via the one or more I/O interfaces, wherein the one or more hardware processors are configured by the instructions to:
• receive an application to be executed in the serverless execution environment;
• and
• perform the workload characterization-based capacity planning of the application using a serverful execution environment, the workload characterization-based capacity planning comprising:
• checking whether a storage requirement of the application, to create an execution environment of the application, is more than a maximum temporary storage available on a serverless instance of the serverless execution environment;
• refactoring the application into a plurality of Application Programming Interfaces (APIs) if the storage requirement is more than the maximum temporary storage, wherein the plurality of APIs are independent of each other and are run on a plurality of serverless instances of the serverless execution environment, wherein a Point to Point (P2P) communication between newly created APIs is performed using an Elastic File System (EFS) enabling the plurality of APIs to read and write data from the EFS, wherein the plurality of APIs pertain to an information extraction and includes an Optical Character Recognition (OCR) API, a Data Type API, a Spatial Relation Creator API, and a Table Detection API;
• checking whether a compute requirement of each of the plurality of APIs is within a maximum number of cores provided by the serverless execution environment;
• in response to determining that the compute requirement of one or more APIs among the plurality of APIs is not satisfied by the maximum number of cores, performing one of:
• a) refactoring the one or more APIs into multiple APIs till the compute requirement is below the maximum number of cores;
• and
• b) launching each thread on a unique serverless instance, if the one or more APIs have a multi-threaded component, and functions are executed in parallel;
• running each data unit from among a plurality of data units in a dataset used by the application for each of the plurality of APIs on the serverful execution environment by varying number of cores provided by a server in the serverful execution environment from a minimum number of cores to the maximum number of cores if the compute requirement is satisfied by the maximum number of cores, wherein the plurality of data units in the dataset comprise a mix of identical data units and unidentical data units for similar type of data;
• determining an average execution time for each data unit from among the plurality of data units of the dataset for each of the plurality of APIs;
• recording an optimal number of cores for each of the plurality of APIs, beyond which there is no further decrease in execution time of each of the plurality of APIs;
• and
• configuring the plurality of serverless instances for executing the application in accordance with the optimal number of cores for each of the plurality of APIs in the serverless execution environment.

One or more non-transitory machine-readable information storage mediums comprising one or more instructions which when executed by one or more hardware processors cause:

• receiving, an application to be executed in the serverless execution environment;
• and
• performing, the workload characterization-based capacity planning of the application using a serverful execution environment, wherein performing the workload characterization-based capacity planning comprising:
• checking whether a storage requirement of the application, to create an execution environment of the application, is more than a maximum temporary storage available on a serverless instance of the serverless execution environment;
• refactoring the application into a plurality of Application Programming Interfaces (APIs) if the storage requirement is more than the maximum temporary storage, wherein the plurality of APIs are independent of each other and run on a plurality of serverless instances of the serverless execution environment, wherein a Point to Point (P2P) communication between the plurality of APIs is performed using an Elastic File System (EFS) enabling the plurality of APIs to read and write data from the EFS, wherein the plurality of APIs pertain to an information extraction and includes an Optical Character Recognition (OCR) API, a Data Type API, a Spatial Relation Creator API, and a Table Detection API;
• checking whether a compute requirement of each of the plurality of APIs is within a maximum number of cores provided by the serverless execution environment;
• in response to determining that the compute requirement of one or more APIs among the plurality of APIs is not satisfied by the maximum number of cores, performing one of:
• a) refactoring the one or more APIs into multiple APIs till the compute requirement is below the maximum number of cores;
• and
• b) launching each thread on a unique serverless instance, if the one or more APIs have a multi-threaded component, and functions are executed in parallel;
• running each data unit from among a plurality of data units in a dataset used by the application for each of the plurality of APIs on the serverful execution environment by varying number of cores provided by a server in the serverful execution environment from a minimum number of cores to the maximum number of cores if the compute requirement is satisfied by the maximum number of cores, wherein the plurality of data units in the dataset comprise a mix of identical data units and unidentical data units for similar type of data;
• determining an average execution time for each data unit from among the plurality of data units of the dataset for each of the plurality of APIs;
• recording an optimal number of cores for each of the plurality of APIs, beyond which there is no further decrease in execution time of each of the plurality of APIs;
• and
• configuring the plurality of serverless instances for executing the application in accordance with the optimal number of cores for each of the plurality of APIs in the serverless execution environment.