arXiv subjects

Xingbo Wu

Publications and source records attributed to Xingbo Wu.

5 recordsLinked to original sources

arXiv2025

Good things come in small packages: Should we build AI clusters with Lite-GPUs?

To match the blooming demand of generative AI workloads, GPU designers have so far been trying to pack more and more compute and memory into single complex and expensive packages. However, there is growing uncertainty about the scalability of individual GPUs and thus AI clusters, as state-of-the-art GPUs are already displaying packaging, yield, and cooling limitations. We propose to rethink the design and scaling of AI clusters through efficiently-connected large clusters of Lite-GPUs, GPUs with single, small dies and a fraction of the capabilities of larger GPUs. We think recent advances in co-packaged optics can enable distributing AI workloads onto many Lite-GPUs through high bandwidth and efficient communication. In this paper, we present the key benefits of Lite-GPUs on manufacturing cost, blast radius, yield, and power efficiency; and discuss systems opportunities and challenges around resource, workload, memory, and network management.

cs.AR↗

arXiv2025

Managed-Retention Memory: A New Class of Memory for the AI Era

AI clusters today are one of the major uses of High Bandwidth Memory (HBM). However, HBM is suboptimal for AI workloads for several reasons. Analysis shows HBM is overprovisioned on write performance, but underprovisioned on density and read bandwidth, and also has significant energy per bit overheads. It is also expensive, with lower yield than DRAM due to manufacturing complexity. We propose a new memory class: Managed-Retention Memory (MRM), which is more optimized to store key data structures for AI inference workloads. We believe that MRM may finally provide a path to viability for technologies that were originally proposed to support Storage Class Memory (SCM). These technologies traditionally offered long-term persistence (10+ years) but provided poor IO performance and/or endurance. MRM makes different trade-offs, and by understanding the workload IO patterns, MRM foregoes long-term data retention and write performance for better potential performance on the metrics important for these workloads.

cs.AR↗

arXiv2021

Efficient Data Management with a Flexible Address Space

Data management applications store their data using structured files in which data are usually sorted to serve indexing and queries. However, in-place insertions and removals of data are not naturally supported in a file's address space. To avoid repeatedly rewriting existing data in a sorted file to admit changes in place, applications usually employ extra layers of indirections, such as mapping tables and logs, to admit changes out of place. However, this approach leads to increased access cost and excessive complexity. This paper presents a novel storage engine that provides a flexible address space, where in-place updates of arbitrary-sized data, such as insertions and removals, can be performed efficiently. With this mechanism, applications can manage sorted data in a linear address space with minimal complexity. Extensive evaluations show that a key-value store built on top of it can achieve high performance and efficiency with a simple implementation.

cs.OS↗

arXiv2021

REMIX: Efficient Range Query for LSM-trees

LSM-tree based key-value (KV) stores organize data in a multi-level structure for high-speed writes. Range queries on traditional LSM-trees must seek and sort-merge data from multiple table files on the fly, which is expensive and often leads to mediocre read performance. To improve range query efficiency on LSM-trees, we introduce a space-efficient KV index data structure, named REMIX, that records a globally sorted view of KV data spanning multiple table files. A range query on multiple REMIX-indexed data files can quickly locate the target key using a binary search, and retrieve subsequent keys in sorted order without key comparisons. We build RemixDB, an LSM-tree based KV-store that adopts a write-efficient compaction strategy and employs REMIXes for fast point and range queries. Experimental results show that REMIXes can substantially improve range query performance in a write-optimized LSM-tree based KV-store.

cs.DB↗

arXiv2019

Wormhole: A Fast Ordered Index for In-memory Data Management

In-memory data management systems, such as key-value stores, have become an essential infrastructure in today's big-data processing and cloud computing. They rely on efficient index structures to access data. While unordered indexes, such as hash tables, can perform point search with O(1) time, they cannot be used in many scenarios where range queries must be supported. Many ordered indexes, such as B+ tree and skip list, have a O(log N) lookup cost, where N is number of keys in an index. For an ordered index hosting billions of keys, it may take more than 30 key-comparisons in a lookup, which is an order of magnitude more expensive than that on a hash table. With availability of large memory and fast network in today's data centers, this O(log N) time is taking a heavy toll on applications that rely on ordered indexes. In this paper we introduce a new ordered index structure, named Wormhole, that takes O(log L) worst-case time for looking up a key with a length of L. The low cost is achieved by simultaneously leveraging strengths of three indexing structures, namely hash table, prefix tree, and B+ tree, to orchestrate a single fast ordered index. Wormhole's range operations can be performed by a linear scan of a list after an initial lookup. This improvement of access efficiency does not come at a price of compromised space efficiency. Instead, Wormhole's index space is comparable to those of B+ tree and skip list. Experiment results show that Wormhole outperforms skip list, B+ tree, ART, and Masstree by up to 8.4x, 4.9x, 4.3x, and 6.6x in terms of key lookup throughput, respectively.

cs.DB↗