Vijayan Prabhakaran

Quincy: fair scheduling for distributed computing clusters

This paper addresses the problem of scheduling concurrent jobs on clusters where application data is stored on the computing nodes. This setting, in which scheduling computations close to their data is crucial for performance, is increasingly common and arises in systems such as MapReduce, Hadoop, and Dryad as well as many grid-computing environments. We argue that data-intensive computation benefits from a fine-grain resource sharing model that differs from the coarser semi-static resource allocations implemented by most existing cluster computing architectures. The problem of scheduling with locality and fairness constraints has not previously been extensively studied under this resource-sharing model. We introduce a powerful and flexible new framework for scheduling concurrent distributed jobs with fine-grain resource sharing. The scheduling problem is mapped to a graph datastructure, where edge weights and capacities encode the competing demands of data locality, fairness, and starvation-freedom, and a standard solver computes the optimal online schedule according to a global cost model. We evaluate our implementation of this framework, which we call Quincy, on a cluster of a few hundred computers using a varied workload of data-and CPU-intensive jobs. We evaluate Quincy against an existing queue-based algorithm and implement several policies for each scheduler, with and without fairness constraints. Quincy gets better fairness when fairness is requested, while substantially improving data locality. The volume of data transferred across the cluster is reduced by up to a factor of 3.9 in our experiments, leading to a throughput increase of up to 40%.

IRON file systems

Commodity file systems trust disks to either work or fail completely, yet modern disks exhibit more complex failure modes. We suggest a new fail-partial failure model for disks, which incorporates realistic localized faults such as latent sector errors and block corruption. We then develop and apply a novel failure-policy fingerprinting framework, to investigate how commodity file systems react to a range of more realistic disk failures. We classify their failure policies in a new taxonomy that measures their Internal RObustNess (IRON), which includes both failure detection and recovery techniques. We show that commodity file system failure policies are often inconsistent, sometimes buggy, and generally inadequate in their ability to recover from partial disk failures. Finally, we design, implement, and evaluate a prototype IRON file system, Linux ixt3, showing that techniques such as in-disk checksumming, replication, and parity greatly enhance file system robustness while incurring minimal time and space overheads.

Consistency-based service level agreements for cloud storage

Choosing a cloud storage system and specific operations for reading and writing data requires developers to make decisions that trade off consistency for availability and performance. Applications may be locked into a choice that is not ideal for all clients and changing conditions. Pileus is a replicated key-value store that allows applications to declare their consistency and latency priorities via consistency-based service level agreements (SLAs). It dynamically selects which servers to access in order to deliver the best service given the current configuration and system conditions. In application-specific SLAs, developers can request both strong and eventual consistency as well as intermediate guarantees such as read-my-writes. Evaluations running on a worldwide test bed with geo-replicated data show that the system adapts to varying client-server latencies to provide service that matches or exceeds the best static consistency choice and server selection scheme.

Differential RAID: Rethinking RAID for SSD reliability

SSDs exhibit very different failure characteristics compared to hard drives. In particular, the bit error rate (BER) of an SSD climbs as it receives more writes. As a result, RAID arrays composed from SSDs are subject to correlated failures. By balancing writes evenly across the array, RAID schemes can wear out devices at similar times. When a device in the array fails towards the end of its lifetime, the high BER of the remaining devices can result in data loss. We propose Diff-RAID, a parity-based redundancy solution that creates an age differential in an array of SSDs. Diff-RAID distributes parity blocks unevenly across the array, leveraging their higher update rate to age devices at different rates. To maintain this age differential when old devices are replaced by new ones, Diff-RAID reshuffles the parity distribution on each drive replacement. We evaluate Diff-RAIDs reliability by using real BER data from 12 flash chips on a simulator and show that it is more reliable than RAID-5, in some cases by multiple orders of magnitude. We also evaluate Diff-RAIDs performance using a software implementation on a 5-device array of 80 GB Intel X25-M SSDs and show that it offers a trade-off between throughput and reliability.

Differential RAID: rethinking RAID for SSD reliability

Deployment of SSDs in enterprise settings is limited by the low erase cycles available on commodity devices. Redundancy solutions such as RAID can potentially be used to protect against the high Bit Error Rate (BER) of aging SSDs. Unfortunately, such solutions wear out redundant devices at similar rates, inducing correlated failures as arrays age in unison. We present Diff-RAID, a new RAID variant that distributes parity unevenly across SSDs to create age disparities within arrays. By doing so, Diff-RAID balances the high BER of old SSDs against the low BER of young SSDs. Diff-RAID provides much greater reliability for SSDs compared to RAID-4 and RAID-5 for the same space overhead, and offers a trade-off curve between throughput and reliability.

Tango: distributed data structures over a shared log

Distributed systems are easier to build than ever with the emergence of new, data-centric abstractions for storing and computing over massive datasets. However, similar abstractions do not exist for storing and accessing meta-data. To fill this gap, Tango provides developers with the abstraction of a replicated, in-memory data structure (such as a map or a tree) backed by a shared log. Tango objects are easy to build and use, replicating state via simple append and read operations on the shared log instead of complex distributed protocols; in the process, they obtain properties such as linearizability, persistence and high availability from the shared log. Tango also leverages the shared log to enable fast transactions across different objects, allowing applications to partition state across machines and scale to the limits of the underlying log without sacrificing consistency.

Improving file system reliability with I/O shepherding

We introduce a new reliability infrastructure for file systems called I/O shepherding. I/O shepherding allows a file system developer to craft nuanced reliability policies to detect and recover from a wide range of storage system failures. We incorporate shepherding into the Linux ext3 file system through a set of changes to the consistency management subsystem, layout engine, disk scheduler, and buffer cache. The resulting file system, CrookFS, enables a broad class of policies to be easily and correctly specified. We implement numerous policies, incorporating data protection techniques such as retry, parity, mirrors, checksums, sanity checks, and data structure repairs; even complex policies can be implemented in less than 100 lines of code, confirming the power and simplicity of the shepherding framework. We also demonstrate that shepherding is properly integrated, adding less than 5% overhead to the I/O path.

Semantically-smart disk systems: past, present, and future

In this paper we describe research that has been on-going within our group for the past four years on semantically-smart disk systems. A semantically-smart system goes beyond typical block-based storage systems by extracting higher-level information from the stream of traffic to disk; doing so enables new and interesting pieces of functionality to be implemented within low-level storage systems. We first describe the development of our efforts over the past four years, highlighting the key technologies needed to build semantically-smart systems as well as the main weaknesses of our approach. We then discuss future directions in the design and implementation of smarter storage systems.

Model-based failure analysis of journaling file systems

We propose a novel method to measure the robustness of journaling file systems under disk write failures. In our approach, we build models of how journaling file systems order disk writes under different journaling modes and use these models to inject write failures during file system updates. Using our technique, we analyze if journaling file systems maintain on-disk consistency in the presence of disk write failures. We apply our technique to three important Linux journaling file systems: ext3, Reiserfs, and IBM JFS. From our analysis, we identify several design flaws and correctness bugs in these file systems, which can cause serious file system errors ranging from data corruption to unmountable file systems.

CORFU: A distributed shared log

CORFU is a global log which clients can append-to and read-from over a network. Internally, CORFU is distributed over a cluster of machines in such a way that there is no single I/O bottleneck to either appends or reads. Data is fully replicated for fault tolerance, and a modest cluster of about 16--32 machines with SSD drives can sustain 1 million 4-KByte operations per second. The CORFU log enabled the construction of a variety of distributed applications that require strong consistency at high speeds, such as databases, transactional key-value stores, replicated state machines, and metadata services.