Prashant Kumar Singh

This patent relates to providing Lambda resellers the ability to take actions on functions via IP addresses associated with their customers. Lambda provisions a range of IP addresses for resellers. The resellers then assign each of these provisioned IP addresses to their customers. The resellers' customers can provide an IP address during a function invoke which Lambda receives. Lambda can execute a function and communicate with the reseller during the function execution and provide an IP… Show more

This patent relates to providing Lambda resellers the ability to take actions on functions via IP addresses associated with their customers. Lambda provisions a range of IP addresses for resellers. The resellers then assign each of these provisioned IP addresses to their customers. The resellers' customers can provide an IP address during a function invoke which Lambda receives. Lambda can execute a function and communicate with the reseller during the function execution and provide an IP address associated with the customer that invoked the function. The reseller can identify the customer associated with the IP address and take actions. Show less

This patent relates to a Serverless Network Interface (SNI) primitive for serverless applications, such as Lambda functions running on micro-virtual machines (micro-vms). Existing solutions often tie networking configuration to each serverless functionality, causing each instance to be created with a static, mirrored configuration. The SNI enables customers to define reusable network connectivity interfaces for their serverless applications. It exists parallel to the Elastic Network Interface… Show more

This patent relates to a Serverless Network Interface (SNI) primitive for serverless applications, such as Lambda functions running on micro-virtual machines (micro-vms). Existing solutions often tie networking configuration to each serverless functionality, causing each instance to be created with a static, mirrored configuration. The SNI enables customers to define reusable network connectivity interfaces for their serverless applications. It exists parallel to the Elastic Network Interface (ENI) but differs in that SNIs are attached to serverless applications backed by multiple VMs, while ENIs are attached to individual compute instances. The SNI serves as a serverless way to define network configurations, including connectivity types, firewall rules (ingress/egress controls), and DNS configurations for serverless platforms like Athena, Glue, and Lambda. Each copy of a serverless function can be configured to attach to the SNI, inheriting its configuration. By decoupling network configuration from the application, SNI enables reusing the same network configuration across multiple applications, reducing the need for duplicating network resources. Show less

This patent relates to generating and managing service virtual computing components that each provide multiple computing services. An application, such as software-as-a-service (SaaS) application or a serverless application, may be composed of multiple service virtual computing components, such as virtual machine instances, containers, and the like. To reduce performance and/or operational issues in the computing environment in which the application is executing (e.g., a host computing device… Show more

This patent relates to generating and managing service virtual computing components that each provide multiple computing services. An application, such as software-as-a-service (SaaS) application or a serverless application, may be composed of multiple service virtual computing components, such as virtual machine instances, containers, and the like. To reduce performance and/or operational issues in the computing environment in which the application is executing (e.g., a host computing device in a data center or cloud environment), computing services with related functionality may be provided by a single service virtual computing component, rather than having each computing service being provided by its own individual virtual computing component. Show less

This patent relates to facilitating direct communication to applications running on an on-demand code execution service (Lambda, App Runner, etc.) from devices external to the on-demand code execution service. Host computing devices (Lambda workers) may be assigned a unique pool of publicly-accessible network addresses based on the network address prefix of the respective host computing devices (e.g., a range of IPv6 addresses all having the same prefix that is uniquely assigned to the host… Show more

This patent relates to facilitating direct communication to applications running on an on-demand code execution service (Lambda, App Runner, etc.) from devices external to the on-demand code execution service. Host computing devices (Lambda workers) may be assigned a unique pool of publicly-accessible network addresses based on the network address prefix of the respective host computing devices (e.g., a range of IPv6 addresses all having the same prefix that is uniquely assigned to the host computing device). When an application instance is launched within a virtual machine (VM) on a host computing device, the VM may be provisioned with a publicly-available network address from the pool assigned to the host computing device. The VM may therefore be accessed from outside the on-demand code execution platform using the provisioned network address. Show less

This patent relates to enabling serverless function invokers, such as Lambda customers, to opt-into using a private path for traffic from their serverless functions to other managed services. The private path uses private IP addresses and ensures that traffic is entirely within a secure managed network infrastructure within which the serverless function and other managed services operate. Use of a such a private path ensures the traffic doesn’t use a public IP address. For example, customers… Show more

This patent relates to enabling serverless function invokers, such as Lambda customers, to opt-into using a private path for traffic from their serverless functions to other managed services. The private path uses private IP addresses and ensures that traffic is entirely within a secure managed network infrastructure within which the serverless function and other managed services operate. Use of a such a private path ensures the traffic doesn’t use a public IP address. For example, customers that opt-in can have traffic from their serverless functions routed through virtual private cloud (VPC) endpoints that are automatically set up and managed as part of the serverless code execution system. The VPC endpoints are configured to route traffic only through the managed network infrastructure and do not have any connectivity to a public IP address. Show less

This patent relates to allowing serverless function owners to access resources in a virtual private cloud (VPC) using user-surface connectors. Rather than handling function-to-VPC communication implicitly, this patent describes how a serverless system can surface explicit objects that a user can directly control to specify permissions, throttles, and scaling limits. This allows for separately addressing functions and connectors, and more clearly conveying to users how functions can communicate… Show more

This patent relates to allowing serverless function owners to access resources in a virtual private cloud (VPC) using user-surface connectors. Rather than handling function-to-VPC communication implicitly, this patent describes how a serverless system can surface explicit objects that a user can directly control to specify permissions, throttles, and scaling limits. This allows for separately addressing functions and connectors, and more clearly conveying to users how functions can communicate to VPCs Show less

This patent relates to supporting execution of tightly coupled parallel applications on a serverless computing system (e.g., AWS Lambda) by implementing a fork/join mechanism that allows a parent task executing on the serverless computing systems to create child tasks and communicate with them via an ephemeral mesh network. An ephemeral cluster manager processes requests from the parent task to create child tasks and creates and maintains an ephemeral mesh network with the parent task and its… Show more

This patent relates to supporting execution of tightly coupled parallel applications on a serverless computing system (e.g., AWS Lambda) by implementing a fork/join mechanism that allows a parent task executing on the serverless computing systems to create child tasks and communicate with them via an ephemeral mesh network. An ephemeral cluster manager processes requests from the parent task to create child tasks and creates and maintains an ephemeral mesh network with the parent task and its child tasks. The ephemeral mesh network allows low-latency communication between tasks executing in different sandboxed environments of the serverless computing system and is removed when the parent task ceases executing. Show less

This patent relates to managing provision of network-stored data sets among instances of serverless function code executing in an on-demand manner (e.g., Lambda). A customer may specify one or more initial data sets to be made available to their serverless functions when the functions begin execution, without specific loading operations required by the functions. For example, the initial data sets are to be made available to the functions via typical local files system data access methods… Show more

This patent relates to managing provision of network-stored data sets among instances of serverless function code executing in an on-demand manner (e.g., Lambda). A customer may specify one or more initial data sets to be made available to their serverless functions when the functions begin execution, without specific loading operations required by the functions. For example, the initial data sets are to be made available to the functions via typical local files system data access methods (e.g., a special “/data” directory). When the functions are invoked, the code execution system can obtain the initial data sets from source data stores and place them in caches that are closer to the host devices on which the functions are executed. When a function reads data from an initial data set using local file system data access methods, the code execution system can load the requested data from the cache to local storage. Moreover, the data sets may, when desired, be made available persistently across function invocations. Show less

This patent relates to managing access to data sets across functions in a serverless environment (e.g., Lambda) such that functions can share data with other functions. An API is provided by which functions can store data sets to be shared with other functions, and by which functions can access data sets shared by other functions. A data set can be generated or obtained by a function, and a sharable data reference can be generated. The API shields developers from the work of moving, storing… Show more

This patent relates to managing access to data sets across functions in a serverless environment (e.g., Lambda) such that functions can share data with other functions. An API is provided by which functions can store data sets to be shared with other functions, and by which functions can access data sets shared by other functions. A data set can be generated or obtained by a function, and a sharable data reference can be generated. The API shields developers from the work of moving, storing, and controlling access to the data sets. In addition, the API allows functions to access large data sets that would not be possible otherwise. For example, data sets may be 100+ GB, which is not able to be fetched and processed by a function due to bandwidth throttling and limitations on execution time. By configuring some functions as “priming functions” that obtain large data sets as subsets, a separate set of “executor functions” can then access and use the large data sets as needed without first retrieving them Show less