Amazon Technologies, Inc.

Ajay Nair

Marc John Brooker

Timothy Allen Wagner

Dylan Chandler Thomas

Craig Wesley Howard

David Alexander Dunlap

Shubham Katiyar

Venkatesh Vijayaraghavan

Manigandan Radhakrishnan

Yilmaz Can Cecen

Vo Vuong

Meenakshi Vembusubramanian

Derek Steven Manwaring

Scott Daniel Wisniewski

Jonathan Paul Thompson

Colm MacCárthaigh

An on-demand code execution environment present in points of presence (POPs) and in regions serviced by the POPs is provided herein. For example, a POP may receive a request to execute a task associated with user-defined code. If the POP determines that the computing resources necessary to execute a received task are not available or that the POP should not execute the received task for another reason (e.g., the task is not commonly received and the computing resources needed to execute the task are therefore best allocated for other requests), the POP can forward the task to a region that the POP services for execution by an on-demand code execution environment present in the region. The on-demand code execution environment present in the region can execute the task and forward the results of the execution to the POP for distribution back to a user device that requested the task execution.

Multi-region request-driven code execution system

Computer systems and methods are disclosed to implement a virtual machine monitor (VMM) that stores cryptographic keys for guest virtual machines (VMs) and securely executes cryptographic operations on the VMs' behalf using the stored cryptographic keys. The cryptographic keys are maintained in a key store that is accessible to the VMM but not accessible to the guest VMs. The cryptographic operations are executed in a manner that does not reveal the cryptographic keys to the guest VMs. In embodiments, the guest VMs may invoke the cryptographic operations via a device driver, a memory access interface, or some other mechanism. Advantageously, the guest VMs cannot obtain the cryptographic keys in their own memory space, so that the keys cannot be exfiltrated from the guest VMs. Embodiments of the VMM may be used to implement cryptographic operations such as request signing and verification, data encryption and decryption, and others.

Virtual machine monitor providing secure cryptographic operations

A system for providing security mechanisms for secure execution of program code is described. The system may be configured to maintain a plurality of virtual machine instances. The system may be further configured to receive a request to execute a program code and allocate computing resources for executing the program code on one of the virtual machine instances. One mechanism involves executing program code according to a user-specified security policy. Another mechanism involves executing program code that may be configured to communicate or interface with an auxiliary service. Another mechanism involves splitting and executing program code in a plurality of portions, where some portions of the program code are executed in association with a first level of trust and some portions of the program code are executed with different levels of trust.

Security protocols for low latency execution of program code

Systems and methods are described for providing escalation-resistant network-accessible services by providing the service through a set of service instances, each executing in an environment with privileges scoped based on a user requesting to access the service. Each service instance can be implemented by code on a serverless code system, executed in response to a user request to access the service. Because the code is executed in an environment with privileges scoped to those of a requesting user, the code itself need not attempt to limit the privileges or a requesting user. For that reason, potential for privilege escalations of the service are reduced, even if vulnerabilities in the code might otherwise allow for such escalations.

Escalation-resistant secure network services using request-scoped authentication information

Systems and methods are described for transforming a data set within a data source into a series of task calls to an on-demand code execution environment. The environment can utilize pre-initialized virtual machine instances to enable execution of user-specified code in a rapid manner, without delays typically caused by initialization of the virtual machine instances, and are often used to process data in near-real time, as it is created. However, limitations in computing resources may inhibit a user from utilizing an on-demand code execution environment to simultaneously process a large, existing data set. The present application provides a task generation system that can iteratively retrieve data items from an existing data set and generate corresponding task calls to the on-demand computing environment. The calls can be ordered to address dependencies of the data items, such as when a first data item depends on prior processing of a second data item.

Dependency-based process of pre-existing data sets at an on demand code execution environment

Systems and methods are described for transforming a data set within a data source into a series of task calls to an on-demand code execution environment or other distributed code execution environment. Such environments utilize pre-initialized virtual machine instances to enable execution of user-specified code in a rapid manner, without delays typically caused by initialization of the virtual machine instances, and are often used to process data in near-real time, as it is created. However, limitations in computing resources may inhibit a user from utilizing an on-demand code execution environment to simultaneously process a large, existing data set. The present application provides a task generation system that can iteratively retrieve data items from an existing data set and generate corresponding task calls to the on-demand computing environment, while ensuring that at least one task call for each data item within the existing data set is made.

Processing pre-existing data sets at an on demand code execution environment

Systems and methods are described for implementing execution guarantees in an on-demand code execution system or other distributed code execution environment, such that the on-demand code execution system attempts to execute code only a desired number of times. The on-demand code execution system can utilize execution identifiers to distinguish between new and duplicative requests, and can decline to allocate computing resources for duplicative requests. The on-demand code execution system can further detect errors during execution, and rollback the execution to undo the execution's effects. The on-demand code execution system can then restart execution until the code has been execute the desired number of times.

Execution guarantees in an on-demand network code execution system

A service manages a plurality of virtual machine instances for low latency execution of user codes. The service can provide the capability to execute user code in response to events triggered on various event sources and initiate execution of other control functions to improve the code execution environment in response to detecting errors or unexpected execution results. The service may maintain or communicate with a separate storage area for storing code execution requests that were not successfully processed by the service. Requests stored in such a storage area may subsequently be re-processed by the service.

Event-driven computing

Systems and methods are described for handling requests to execute idempotent code in an on-demand code execution system or other distributed code execution environment. Idempotent code can generally include code that produces the same outcome even when executed multiple times, so long as dependencies for the code are in the same state as during a prior execution. Due to this feature, multiple executions of idempotent code may inefficiently use computing resources, particularly in on-demand code execution system (which may require, for example, generation and provisioning of an appropriate execution environment for the code). Aspects of the present disclosure enable the on-demand code execution system to process requests to execute code by verifying whether dependency states associated with the code have changed since a prior execution. If dependency states have not changed, no execution need occur, and the overall computing resource us of the on-demand code execution system is decreased.

Idempotent task execution in on-demand network code execution systems

An on-demand code execution environment present in points of presence (POPs) and in regions serviced by the POPs is provided herein. For example, a POP may receive a request to execute a task associated with user-defined code. If the POP determines that the computing resources necessary to execute a received task are not available or that the POP should not execute the received task for another reason (e.g., the task is not commonly received and the computing resources needed to execute the task are therefore best allocated for other requests), the POP can identify a region suitable for executing the task and forward the task to the identified region. An on-demand code execution environment present in the identified region can execute the task and forward the results of the execution to the POP for distribution back to a user device that requested the task execution.

Efficient region selection system for executing request-driven code

Systems and methods are described for managing asynchronous code executions in an on-demand code execution system or other distributed code execution environment, in which multiple execution environments, such as virtual machine instances, can be used to enable rapid execution of user-submitted code. When asynchronous executions occur, a first execution may call a second execution, but not immediately need the second execution to complete. To efficiently allocate computing resources, this disclosure enables the second execution to be scheduled accordingly to a state of the on-demand code execution system, while still ensuring the second execution completes prior to the time required by the first execution. Scheduling of executions can, for example, enable more efficient load balancing on the on-demand code execution system.

Asynchronous task management in an on-demand network code execution environment

Systems and methods are described for managing cross-account access to tasks on an on-demand code execution environment or other distributed code execution environment. Such environments utilize pre-initialized virtual machine instances to enable execution of user-specified code in a rapid manner, without delays typically caused by initialization of the virtual machine instances. However, to ensure security, the code of different users is generally maintained separately, and executed on separate virtual machines. Embodiments described herein enable users of a first account to execute code of a second account, without gaining access to the code itself and while maintaining the privacy and security of each account. Specifically, aliases for a task of a first account can be created on a task of a second account, and used to invoke that task on behalf of the first account. Aliases may also allow users to customize how the task is executed.

On-demand network code execution with cross-account aliases

Systems and methods are described for utilizing cross-account access to tasks on an on-demand code execution environment or other distributed code execution environment to implement an application programming interface (API) on a network-accessible service. An on-demand code execution environment can utilize pre-initialized virtual machine instances to enable execution of user-specified code in a rapid manner, without delays typically caused by initialization of the virtual machine instances. While users may generally execute their own code, the present application enables code of a first user to be executed by a second user, while maintaining the privacy and security of the code and associated accounts. Further, the present application enables a set of tasks to be grouped together as an API, enabling any user to create an API for a service, while the on-demand code execution environment manages execution of the tasks and allocation of underlying computing resources.

Growing community of inventors

Seattle, WA, United States of America

Ajay Nair