Search for: All records

Finding the connected components of a graph is a fundamental prob- lem with uses throughout computer science and engineering. The task of computing connected components becomes more difficult when graphs are very large, or when they are dynamic, meaning the edge set changes over time subject to a stream of edge inser- tions and deletions. A natural approach to computing the connected components on a large, dynamic graph stream is to buy enough RAM to store the entire graph. However, the requirement that the graph fit in RAM is prohibitive for very large graphs. Thus, there is an unmet need for systems that can process dense dynamic graphs, especially when those graphs are larger than available RAM. We present a new high-performance streaming graph-processing system for computing the connected components of a graph. This system, which we call GraphZeppelin, uses new linear sketching data structures (CubeSketch) to solve the streaming connected components problem and as a result requires space asymptotically smaller than the space required for a lossless representation of the graph. GraphZeppelin is optimized for massive dense graphs: GraphZeppelin can process millions of edge updates (both inser- tions and deletions) per second, even when the underlying graph is far too large to fit in available RAM. As a result GraphZeppelin vastly increases the scale of graphs that can be processed. 

A data structure A is said to be dynamically optimal over a class of data structures C if A is constant- competitive with every data structure C ∈ C. Much of the research on binary search trees in the past forty years has focused on studying dynamic optimality over the class of binary search trees that are modified via rotations (and indeed, the question of whether splay trees are dynamically optimal has gained notoriety as the so-called dynamic-optimality conjecture). Recently, researchers have extended this to consider dynamic optimality over certain classes of external-memory search trees. In particular, Demaine, Iacono, Koumoutsos, and Langerman propose a class of external-memory trees that support a notion of tree rotations, and then give an elegant data structure, called the Belga B-tree, that is within an O(log log N )-factor of being dynamically optimal over this class. In this paper, we revisit the question of how dynamic optimality should be defined in external memory. A defining characteristic of external-memory data structures is that there is a stark asymmetry between queries and inserts/updates/deletes: by making the former slightly asymptotically slower, one can make the latter significantly asymptotically faster (even allowing for operations with sub-constant amortized I/Os). This asymmetry makes it so that rotation-based search trees are not optimal (or even close to optimal) in insert/update/delete-heavy external-memory workloads. To study dynamic optimality for such workloads, one must consider a different class of data structures. The natural class of data structures to consider are what we call buffered-propagation trees. Such trees can adapt dynamically to the locality properties of an input sequence in order to optimize the interactions between different inserts/updates/deletes and queries. We also present a new form of beyond-worst-case analysis that allows for us to formally study a continuum between static and dynamic optimality. Finally, we give a novel data structure, called the Jεllo Tree, that is statically optimal and that achieves dynamic optimality for a large natural class of inputs defined by our beyond-worst-case analysis. 

For nearly six decades, the central open question in the study of hash tables has been to determine the optimal achievable tradeoff curve between time and space. State-of-the-art hash tables offer the following guarantee: If keys/values are Θ(logn) bits each, then it is possible to achieve constant-time insertions/deletions/queries while wasting only O(loglogn) bits of space per key when compared to the information-theoretic optimum. Even prior to this bound being achieved, the target of O(log log n) wasted bits per key was known to be a natural end goal, and was proven to be optimal for a number of closely related problems (e.g., stable hashing, dynamic retrieval, and dynamically-resized filters). This paper shows that O(log log n) wasted bits per key is not the end of the line for hashing. In fact, for any k ∈ [log∗ n], it is possible to achieve O(k)-time insertions/deletions, O(1)-time queries, and O(log(k) n) = Ologlog···logn 􏰟 􏰞􏰝 􏰠 k wasted bits per key (all with high probability in n). This means that, each time we increase inser- tion/deletion time by an additive constant, we reduce the wasted bits per key exponentially. We further show that this tradeoff curve is the best achievable by any of a large class of hash tables, including any hash table designed using the current framework for making constant-time hash tables succinct. Our results hold not just for fixed-capacity hash tables, but also for hash tables that are dynamically resized (this is a fundamental departure from what is possible for filters); and for hash tables that store very large keys/values, each of which can be up to no(1) bits (this breaks with the conventional wisdom that larger keys/values should lead to more wasted bits per key). For very small keys/values, we are able to tighten our bounds to o(1) wasted bits per key, even when k = O(1). Building on this, we obtain a constant-time dynamic filter that uses n􏰕logε−1􏰖+nloge+o(n) bits of space for a wide choice of 

Today’s filters, such as quotient, cuckoo, and Morton, have a trade-off between space and speed; even when moderately full (e.g., 50%-75% full), their performance degrades nontrivially. The result is that today’s systems designers are forced to choose between speed and space usage. In this paper, we present the vector quotient filter (VQF). Locally, the VQF is based on Robin Hood hashing, like the quotient filter, but uses power-of-two-choices hashing to reduce the variance of runs, and thus others consistent, high throughput across load factors. Power-of-two-choices hashing also makes it more amenable to concurrent updates, compared to the cuckoo filter and variants. Finally, the vector quotient filter is designed to exploit SIMD instructions so that all operations have $ (1) cost, independent of the size of the filter or its load factor. We show that the vector quotient flter is 2x faster for inserts compared to the Morton filter (a cuckoo filter variant and state-of- the-art for inserts) and has similar lookup and deletion performance as the cuckoo filter (which is fastest for queries and deletes), despite having a simpler design and implementation. The vector quotient filter has minimal performance decline at high load factors, a problem that has plagued modern filters, including quotient, cuckoo, and Morton. Furthermore, we give a thread-safe version of the vector quotient filter and show that insertion throughput scales 3x with four threads compared to a single thread. 

For any forest G = (V, E) it is possible to orient the edges E so that no vertex in V has out-degree greater than 1. This paper considers the incremental edge-orientation problem, in which the edges E arrive over time and the algorithm must maintain a low-out-degree edge orientation at all times. We give an algorithm that maintains a maximum out-degree of 3 while flipping at most O(log log n) edge orientations per edge insertion, with high probability in n. The algorithm requires worst-case time O(log n log log n) per insertion, and takes amortized time O(1). The previous state of the art required up to O(log n/ log log n) edge flips per insertion. We then apply our edge-orientation results to the problem of dynamic Cuckoo hashing. The problem of designing simple families H of hash functions that are compatible with Cuckoo hashing has received extensive attention. These families H are known to satisfy static guarantees, but do not come typically with dynamic guarantees for the running time of inserts and deletes. We show how to transform static guarantees (for 1-associativity) into near-state-of-the-art dynamic guarantees (for O(1)-associativity) in a black-box fashion. Rather than relying on the family H to supply randomness, as in past work, we instead rely on randomness within our table-maintenance algorithm. 

The binary-forking model is a parallel computation model, formally defined by Blelloch et al., in which a thread can fork a concurrent child thread, recursively and asynchronously. The model incurs a cost of Theta(log n) to spawn or synchronize n tasks or threads. The binary-forking model realistically captures the performance of parallel algorithms implemented using modern multithreaded programming languages on multicore shared-memory machines. In contrast, the widely studied theoretical PRAM model does not consider the cost of spawning and synchronizing threads, and as a result, algorithms achieving optimal performance bounds in the PRAM model may not be optimal in the binary-forking model. Often, algorithms need to be redesigned to achieve optimal performance bounds in the binary-forking model and the non-constant synchronization cost makes the task challenging. In this paper, we show that in the binary-forking model we can achieve optimal or near-optimal span with negligible or no asymptotic blowup in work for comparison-based sorting, Strassen's matrix multiplication (MM), and the Fast Fourier Transform (FFT). Our major results are as follows: (1) A randomized comparison-based sorting algorithm with optimal O(log n) span and O(nlog n) work, both w.h.p. in n. (2) An optimal O(log n) span algorithm for Strassen's matrix multiplication (MM) with only a loglog n - factor blow-up in work as well as a near-optimal O(log n loglog log n) span algorithm with no asymptotic blow-up in work. (3) A near-optimal O(log n logloglog n) span Fast Fourier Transform (FFT) algorithm with less than a log n-factor blow-up in work for all practical values of n (i.e., n le 10 ^10,000) 

First introduced in 1954, the linear-probing hash table is among the oldest data structures in computer science, and thanks to its unrivaled data locality, linear probing continues to be one of the fastest hash tables in practice. It is widely believed and taught, however, that linear probing should never be used at high load factors; this is because of an effect known as primary clustering which causes insertions at a load factor of 1 - 1/x to take expected time O(x2) (rather than the intuitive running time of ⇥(x)). The dangers of primary clustering, first discovered by Knuth in 1963, have now been taught to generations of computer scientists, and have influenced the design of some of the most widely used hash tables in production. We show that primary clustering is not the foregone conclusion that it is reputed to be. We demonstrate that seemingly small design decisions in how deletions are implemented have dramatic effects on the asymptotic performance of insertions: if these design decisions are made correctly, then even if a hash table operates continuously at a load factor of 1 - (1/x), the expected amortized cost per insertion/deletion is O(x). This is because the tombstones left behind by deletions can actually cause an anti-clustering effect that combats primary clustering. Interestingly, these design decisions, despite their remarkable effects, have historically been viewed as simply implementation-level engineering choices. We also present a new variant of linear probing (which we call graveyard hashing) that completely eliminates primary clustering on any sequence of operations: if, when an operation is performed, the current load factor is 1 1/x for some x, then the expected cost of the operation is O(x). Thus we can achieve the data locality of traditional linear probing without any of the disadvantages of primary clustering. One corollary is that, in the external-memory model with a data blocks of size B, graveyard hashing offers the following remarkably strong guarantee: at any load factor 1 1/x satisfying x = o(B), graveyard hashing achieves 1 + o(1) expected block transfers per operation. In contrast, past external-memory hash tables have only been able to offer a 1 + o(1) guarantee when the block size B is at least O(x2). Our results come with actionable lessons for both theoreticians and practitioners, in particular, that well- designed use of tombstones can completely change the asymptotic landscape of how the linear probing behaves (and even in workloads without deletions). 

Making logical copies, or clones, of files and directories is critical to many real-world applications and work- flows, including backups, virtual machines, and containers. An ideal clone implementation meets the follow- ing performance goals: (1) creating the clone has low latency; (2) reads are fast in all versions (i.e., spatial locality is always maintained, even after modifications); (3) writes are fast in all versions; (4) the overall sys- tem is space efficient. Implementing a clone operation that realizes all four properties, which we call a nimble clone, is a long-standing open problem. This article describes nimble clones in B-ε-tree File System (BetrFS), an open-source, full-path-indexed, and write-optimized file system. The key observation behind our work is that standard copy-on-write heuristics can be too coarse to be space efficient, or too fine-grained to preserve locality. On the other hand, a write- optimized key-value store, such as a Bε -tree or an log-structured merge-tree (LSM)-tree, can decouple the logical application of updates from the granularity at which data is physically copied. In our write-optimized clone implementation, data sharing among clones is only broken when a clone has changed enough to warrant making a copy, a policy we call copy-on-abundant-write. We demonstrate that the algorithmic work needed to batch and amortize the cost of BetrFS clone operations does not erode the performance advantages of baseline BetrFS; BetrFS performance even improves in a few cases. BetrFS cloning is efficient; for example, when using the clone operation for container creation, BetrFSoutperforms a simple recursive copy by up to two orders-of-magnitude and outperforms file systems that have specialized Linux Containers (LXC) backends by 3–4×. 

The classical paging problem, introduced by Sleator and Tarjan in 1985, formalizes the problem of caching pages in RAM in order to minimize IOs. Their online formulation ignores the cost of address translation: programs refer to data via virtual addresses, and these must be translated into physical locations in RAM. Although the cost of an individual address translation is much smaller than that of an IO, every memory access involves an address translation, whereas IOs can be infrequent. In practice, one can spend money to avoid paging by over-provisioning RAM; in contrast, address translation is effectively unavoidable. Thus address-translation costs can sometimes dominate paging costs, and systems must simultane- ously optimize both. To mitigate the cost of address translation, all modern CPUs have translation lookaside buffers (TLBs), which are hardware caches of common address translations. What makes TLBs interesting is that a single TLB entry can potentially encode the address translation for many addresses. This is typically achieved via the use of huge pages, which translate runs of contiguous virtual addresses to runs of contiguous physical addresses. Huge pages reduce TLB misses at the cost of increasing the IOs needed to maintain contiguity in RAM. This tradeoff between TLB misses and IOs suggests that the classical paging problem does not tell the full story. This paper introduces the Address-Translation Problem, which formalizes the problem of maintaining a TLB, a page table, and RAM in order to minimize the total cost of both TLB misses and IOs. We present an algorithm that achieves the benefits of huge pages for TLB misses without the downsides of huge pages for IOs. 

Storage devices have complex performance profiles, including costs to initiate IOs (e.g., seek times in hard 15 drives), parallelism and bank conflicts (in SSDs), costs to transfer data, and firmware-internal operations. The Disk-access Machine (DAM) model simplifies reality by assuming that storage devices transfer data in blocks of size B and that all transfers have unit cost. Despite its simplifications, the DAM model is reasonably accurate. In fact, if B is set to the half-bandwidth point, where the latency and bandwidth of the hardware are equal, then the DAM approximates the IO cost on any hardware to within a factor of 2. Furthermore, the DAM model explains the popularity of B-trees in the 1970s and the current popularity of Bε -trees and log-structured merge trees. But it fails to explain why some B-trees use small nodes, whereas all Bε -trees use large nodes. In a DAM, all IOs, and hence all nodes, are the same size. In this article, we show that the affine and PDAM models, which are small refinements of the DAM model, yield a surprisingly large improvement in predictability without sacrificing ease of use. We present benchmarks on a large collection of storage devices showing that the affine and PDAM models give good approximations of the performance characteristics of hard drives and SSDs, respectively. We show that the affine model explains node-size choices in B-trees and Bε -trees. Furthermore, the models predict that B-trees are highly sensitive to variations in the node size, whereas Bε -trees are much less sensitive. These predictions are born out empirically. Finally, we show that in both the affine and PDAM models, it pays to organize data structures to exploit varying IO size. In the affine model, Bε -trees can be optimized so that all operations are simultaneously optimal, even up to lower-order terms. In the PDAM model, Bε -trees (or B-trees) can be organized so that both sequential and concurrent workloads are handled efficiently. We conclude that the DAM model is useful as a first cut when designing or analyzing an algorithm or data structure but the affine and PDAM models enable the algorithm designer to optimize parameter choices and fill in design details.