QFaaS: accelerating and securing serverless cloud networks with QUIC
QFaaS: accelerating and securing serverless cloud networks with QUIC
复制标题
DOI：
10.1145/3542929.3563458
复制
发表时间：
2022-11
期刊：
影响因子：
--
通讯作者：
Kaiyu Hou;Sen-Shou Lin;Yan Chen;V. Yegneswaran
中科院分区：
文献类型：
--
作者：
Kaiyu Hou;Sen-Shou Lin;Yan Chen;V. Yegneswaran
无服务器计算极大地简化了云编程。它将云租户从各种系统管理和资源管理任务中解放出来，例如配置和配置。在这种新的云计算模式下，单个单一的应用程序被划分为单独的无状态功能，即功能即服务(FAAS)，然后将这些功能编排在一起以支持复杂的业务逻辑。但这种增强的灵活性有一个基本成本。函数之间的内部网络连接现在频繁启动，以支持敏捷自动伸缩和函数链等无服务器功能，从而增加了通信延迟。为了降低这一成本，当前的无服务器提供商牺牲了安全性来换取性能，使内部功能通信保持不加密。我们相信，新出现的Quic协议可以为这一挑战提供解决方案，该协议保护并加速了广域内的HTTP通信。我们设计了一个基于Quic的FAAS框架QFaaS，并在OpenFaaS平台上进行了实现。我们的设计明确地确保现有的无服务器应用程序可以直接受益于QFaaS，而无需修改任何应用程序代码。在合成函数和实际应用中的实验表明，QFaaS可以使单个函数和函数链的通信延迟分别减少28%和40%，最终用户响应时间最多节省50ms。
Serverless computing has greatly simplified cloud programming. It liberates cloud tenants from various system administration and resource management tasks, such as configuration and provisioning. Under this new cloud computing paradigm, a single monolithic application is divided into separate stateless functions, i.e., function-as-a-service (FaaS), which are then orchestrated together to support complex business logic. But there is a fundamental cost associated with this enhanced flexibility. Internal network connections between functions are now initiated frequently, to support serverless features such as agile autoscaling and function chains, raising communication latency. To alleviate this cost, current serverless providers sacrifice security for performance, keeping internal function communications unencrypted. We believe that the emerging QUIC protocol, which has secured and accelerated HTTP communications in the wide area, could proffer a solution to this challenge. We design a QUIC-based FaaS framework, called QFaaS, and implement it on the OpenFaaS platform. Our design explicitly ensures that existing serverless applications can directly benefit from QFaaS without any application code modification. Experiments on synthetic functions and real-world applications demonstrate that QFaaS can reduce communication latency for single functions and function chains by 28% and 40%, respectively, and save up to 50 ms in end-user response time.