No abstract

No abstract

Increased access to space has led to an increase in the usage of commodity processors in radiation environments. These processors are vulnerable to transient faults such as single event upsets that may cause bit-flips in processor components. Caches in particular are vulnerable due to their relatively large area, yet are often omitted from fault injection testing because many processors do not provide direct access to cache contents and they are often not fully modeled by simulators. The performance benefits of caches make disabling them undesirable, and the presence of error correcting codes is insufficient to correct for increasingly common multiple bit upsets. This work explores building a program’s cache profile by collecting cache usage information at an instruction granularity via commonly available on-chip debugging interfaces. The profile provides a tighter bound than cache utilization for cache vulnerability estimates (50% for several benchmarks). This can be applied to reduce the number of fault injections required to characterize behavior by at least two-thirds for the benchmarks we examine. The profile enables future work in hardware fault injection for caches that avoids the biases of existing techniques.

Today, distributed systems are typically designed to be largely asynchronous. Designers assume that the network can drop or significantly delay messages at unpredictable times, that there is no way to know how quickly a node might process a message, or how soon it might respond, and that the clocks of different nodes are at most loosely synchronized. These assumptions are certainly safe, but they come at a price: many applications really do need predictable performance, which, on top of an asynchronous system, has to be approximated at great cost and with lots of redundancy, and many distributed protocols for asynchronous systems are much more complex and expensive than their synchronous counterparts. The goal of this paper is to start a discussion about whether the asynchronous model is (still) the right choice for most distributed systems, especially ones that belong to a single administrative domain. We argue that 1) by using ideas from the CPS domain, it is technically feasible to build datacenterscale systems that are fully synchronous, and that 2) such systems would have several interesting advantages over current designs.

No abstract