Modern disk drives read-ahead data and reorder incoming requests in a workload-dependent fashion. This improves their performance, but makes simple analytical models of them inadequate for performance prediction, capacity planning, workload balancing, and so on. To address this problem we have developed a new analytic model for disk drives that do readahead and request reordering. We did so by developing performance models of the disk drive components (queues, caches, and the disk mechanism) and a workload transformation technique for composing them. Our model includes the effects of workload-specific parameters such as request size and spatial locality.The result is capable of predicting the behavior of a variety of real-world devices to within 17% across a variety of workloads and disk drives.
MotivationThere are many reasons for wanting analytical performance models of disk drives and other storage devices. In our case, we were working to develop an automatic attribute-managed storage system that takes in descriptions of the storage workload and automatically designs and configures a storage system that meets those needs [ll]. One component of the solution was an assignment engine that explored the design space -which workload element to assign to which storage device. Each trial in this search required a performance prediction, which meant that we needed a performance model that was both fast (a few milliseconds to execute), accurate (within 30% of the real device), and capable of representing a variety of storage devices.Although much effort was expended in producing analytical models of disk drives in the 1960s and 1970s most recent modeling work has been concentrated on disk arrays, so advances in disk drives such as on-board controllers that cache requests and do readahead and write-behind have largely been ignored. As [25] showed, accurate modeling of these behaviors is important:ignoring caching effects can result in performance prediction errors of over 100%.Additionally, since we needed to match workloads to devices to produce good assignments, the performance model *Lucent/Bell Labs.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.