On large-scale high performance computing (HPC) systems, applications are provisioned with aggregated resources to meet their peak demands for brief periods. This results in resource underutilization because application requirements vary a lot during execution. This problem is particularly pronounced for deep learning applications that are running on leadership HPC systems with a large pool of burst buffers in the form of flash or non-volatile memory (NVM) devices. In this paper, we examine the I/O patterns of deep neural networks and reveal their critical need of loading many small samples randomly for successful training. We have designed a specialized Deep Learning File System (DLFS) that provides a thin set of APIs. Particularly, we design the metadata management of DLFS through an in-memory tree-based sample directory and its file services through the user-level SPDK protocol that can disaggregate the capabilities of NVM Express (NVMe) devices to parallel training tasks. Our experimental results show that DLFS can dramatically improve the throughput of training for deep neural networks on NVMe over Fabric, compared with the kernel-based Ext4 file system. Furthermore, DLFS achieves efficient user-level storage disaggregation with very little CPU utilization.
more »
« less
XRP: In-Kernel Storage Functions with eBPF
With the emergence of microsecond-scale NVMe storage devices, the Linux kernel storage stack overhead has become significant, almost doubling access times. We present XRP, a framework that allows applications to execute user-defined storage functions, such as index lookups or aggregations, from an eBPF hook in the NVMe driver, safely bypassing most of the kernel’s storage stack. To preserve file system semantics, XRP propagates a small amount of kernel state to its NVMe driver hook where the user-registered eBPF functions are called. We show how two key-value stores, BPF-KV, a simple B+-tree key-value store, and WiredTiger, a popular log-structured merge tree storage engine, can leverage XRP to significantly improve throughput and latency.
more »
« less
- Award ID(s):
- 1750558
- NSF-PAR ID:
- 10409936
- Date Published:
- Journal Name:
- Proceedings of the 16th USENIX Symposium on Operating Systems Design and Implementation
- Page Range / eLocation ID:
- 375-393
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Modern NVMe solid state drives offer significantly higher bandwidth and lower latency than prior storage devices. Cur- rent key-value stores struggle to fully utilize the bandwidth of such devices. This paper presents SplinterDB, a new key- value store explicitly designed for NVMe solid-state-drives. SplinterDB is designed around a novel data structure (the STBe-tree) that exposes I/O and CPU concurrency and re- duces write amplification without sacrificing query perfor- mance. STBe-tree combines ideas from log-structured merge trees and Be-trees to reduce write amplification and CPU costs of compaction. The SplinterDB memtable and cache are designed to be highly concurrent and to reduce cache misses. We evaluate SplinterDB on a number of micro- and macro-benchmarks, and show that SplinterDB outperforms RocksDB, a state-of-the-art key-value store, by a factor of 6–10⇥ on insertions and 2–2.6⇥ on point queries, while matching RocksDB on small range queries. Furthermore, SplinterDB reduces write amplification by 2⇥ compared to RocksDB.more » « less
-
Modern NVMe solid state drives offer significantly higher bandwidth and low latency than prior storage devices. Current key-value stores struggle to fully utilize the bandwidth of such devices. This paper presents SplinterDB, a new key-value store explicitly designed for NVMe solid state drives. SplinterDB is designed around a novel data structure (the STBε-tree), that exposes I/O and CPU concurrency and reduces write amplification without sacrificing query performance. STBε-tree combines ideas from log-structured merge trees and Bε-trees to reduce write amplification and CPU costs of compaction. The SplinterDB memtable and cache are designed to be highly concurrent and to reduce cache misses. We evaluate SplinterDB on a number of micro- and macro-benchmarks, and show that SplinterDB outperforms RocksDB, a state-of-the-art key-value store, by a factor of 6–10x on insertions and 2–2.6x on point queries, while matching RocksDB on small range queries. Furthermore, SplinterDB reduces write amplification by 2x compared to RocksDB.more » « less
-
Modern NVMe solid state drives offer significantly higher bandwidth and low latency than prior storage devices. Current key-value stores struggle to fully utilize the bandwidth of such devices. This paper presents SplinterDB, a new key-value store explicitly designed for NVMe solid state drives. SplinterDB is designed around a novel data structure (the STBε-tree), that exposes I/O and CPU concurrency and reduces write amplification without sacrificing query performance. STBε-tree combines ideas from log-structured merge trees and Bε-trees to reduce write amplification and CPU costs of compaction. The SplinterDB memtable and cache are designed to be highly concurrent and to reduce cache misses. We evaluate SplinterDB on a number of micro- and macro-benchmarks, and show that SplinterDB outperforms RocksDB, a state-of-the-art key-value store, by a factor of 6–10x on insertions and 2–2.6x on point queries, while matching RocksDB on small range queries. Furthermore, SplinterDB reduces write amplification by 2x compared to RocksDB.more » « less
-
Traditional network resident functions (e.g., firewalls, network address translation) and middleboxes (caches, load balancers) have moved from purpose-built appliances to software-based components. However, L2/L3 network functions (NFs) are being implemented on Network Function Virtualization (NFV) platforms that extensively exploit kernel-bypass technology. They often use DPDK for zero-copy delivery and high performance. On the other hand, L4/L7 middleboxes, which usually require full network protocol stack support, take advantage of a full-fledged kernel-based system with a greater emphasis on functionality. Thus, L2/L3 NFs and middleboxes continue to be handled by distinct platforms on different nodes.This paper proposes MiddleNet that seeks to overcome this dichotomy by developing a unified network resident function framework that supports L2/L3 NFs and L4/L7 middleboxes. MiddleNet supports function chains that are essential in both NFV and middlebox environments. MiddleNet uses DPDK for zero-copy packet delivery without interrupt-based processing, to enable the ‘bump-in-the-wire’ L2/L3 processing performance required of NFV. To support L4/L7 middlebox functionality, MiddleNet utilizes a consolidated, kernel-based protocol stack processing, avoiding a dedicated protocol stack for each function. MiddleNet fully exploits the event-driven capabilities provided by the extended Berkeley Packet Filter (eBPF) and seamlessly integrates it with shared memory for high-performance communication in L4/L7 middlebox function chains. The overheads for MiddleNet are strictly load-proportional, without needing the dedicated CPU cores of DPDK-based approaches. MiddleNet supports flow-dependent packet processing by leveraging Single Root I/O Virtualization (SR-IOV) to dynamically select packet processing needed (Layer 2 to Layer 7). Our experimental results show that MiddleNet can achieve high performance in such a unified environment.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- The DOI is not currently available.
