CXL Shared Memory Programming: Barely Distributed and Almost Persistent

虽然Compute Express Link（CXL）支持多个节点之间的缓存一致共享内存，但它也引入了新类型的故障——数据可能在进程之前失败，或者进程可能在数据之前失败。缺乏CXL基于共享内存的故障模型使得理解和缓解这些故障具有挑战性。 为了解决这些挑战，本文描述了一种对CXL基于共享内存的故障进行分类和处理的模型：数据和进程故障。 CXL基于共享内存的数据故障会使当前运行的应用程序无法访问或一致性不足。我们认为这种故障不像分布式存储系统中的数据故障，并需要CXL特定的处理。为了解决这个问题，我们研究了传统的数据故障缓解技术，如纠删码和复制，并提出了新的解决方案，以更好地处理CXL基于共享内存系统中的数据故障。接下来，我们研究了进程故障，并将故障和潜在解决方案与PMEM的故障模型和编程解决方案进行了比较。我们认为，尽管PMEM具有一些CXL的特性，但它并未完全解决CXL的易失性和低访问延迟。最后，受到PMEM编程解决方案的启发，我们提出了处理这些新故障的技术。 因此，本文是第一篇定义CXL基于共享内存故障模型并提出针对CXL基于系统特定挑战的定制解决方案的工作。

CXL-based shared memory introduces new types of failures, making it challenging to understand and mitigate these failures. This paper aims to define the CXL-based shared memory failure model and propose tailored solutions to address challenges specific to CXL-based systems.

The paper categorizes and handles the CXL-based shared memory's failures: data and process failures. It proposes new solutions to better handle data failures in CXL-based shared memory systems and techniques to handle process failures, taking inspiration from PMEM programming solutions.

The paper is the first work to define the CXL-based shared memory failure model and propose tailored solutions. It compares the failures and potential solutions with PMEM's failure model and programming solutions. The experiments conducted are not mentioned in the abstract. The paper proposes new solutions to better handle data failures in CXL-based shared memory systems and techniques to handle process failures.

The abstract does not mention any related research in the field.