Lessons learned from the early performance evaluation of Intel optane DC persistent memory in DBMS

Wu, Yinjun; Park, Kwanghyun; Sen, Rathijit; Kroth, Brian; Do, Jaeyoung

doi:10.1145/3399666.3399898

Cited by 24 publications

(8 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Prefetching is done by the iMC and processing is interleaved with I/O very efficiently, making the working set of almost all the queries to fit in the L4 cache, making essentially the two modes running under the same hardware. This comes in contrast with the findings of Wu et al [41], where the Memory mode was slower than the Default mode. However, the authors of this paper perform warm runs, whereas in our case we clear all the caches before performing measurements.…”

Section: Concurrent and Double-write Buffercontrasting

confidence: 99%

“…Complementary to that, Wu et. al [41] have run microbenchmarks, TPC-H, and TPC-C on SQL server 2019 using PMEM in Memory and AppDirect mode (with and without fsdax) and they present high-level conclusions on how DBMS can leverage PMEM. Additionally, Renen et al [38] measure the bandwidth and latency of Intel Optane Persistent Memory and based on their findings they tune log writing and block flushing.…”

Section: Related Workmentioning

confidence: 99%

“…Additionally, databases have many knobs that can be altered, which can affect performance significantly. Although the database community has started to evaluated PMEM on relational workloads [22,25,41], these studies give only part of the picture as they provide only highlevel conclusions and are mostly based on micro-or small-scale benchmarks. Furthermore, they do not experiment with different database knobs and they do not correlate database characteristics with PMEM behaviour.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

How to use Persistent Memory in your Database

Koutsoukos¹,

Bhartia²,

Klimovic³

et al. 2021

Preprint

View full text Add to dashboard Cite

Persistent or Non Volatile Memory (PMEM or NVM) has recently become commercially available under several configurations with different purposes and goals. Despite the attention to the topic, we are not aware of a comprehensive empirical analysis of existing relational database engines under different PMEM configurations. Such a study is important to understand the performance implications of the various hardware configurations and how different DB engines can benefit from them. To this end, we analyze three different engines (PostgreSQL, MySQL, and SQLServer) under common workloads (TPC-C and TPC-H) with all possible PMEM configurations supported by Intel's Optane NVM devices (PMEM as persistent memory in AppDirect mode and PMEM as volatile memory in Memory mode). Our results paint a complex picture and are not always intuitive due to the many factors involved. Based on our findings, we provide insights on how the different engines behave with PMEM and which configurations and queries perform best. Our results show that using PMEM as persistent storage usually speeds up query execution, but with some caveats as the I/O path is not fully optimized. Additionally, using PMEM in Memory mode does not offer any performance advantage despite the larger volatile memory capacity. Through the extensive coverage of engines and parameters, we provide an important starting point for exploiting PMEM in databases and tuning relational engines to take advantage of this new technology.

show abstract

Section: Concurrent and Double-write Buffercontrasting

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

How to use Persistent Memory in your Database

Koutsoukos¹,

Bhartia²,

Klimovic³

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…Other research groups have also tested DCPMMs from an application perspective. However, rather than focusing on high throughput applications, the typical use case of DCPMMs that is found in literature is to accelerate, for example, database workloads [14] or perform faster graph analytics [15]. I summarizes the specification of the machine node used for the evaluation.…”

Section: Several Research Institutes Have Already Tested the Intel®mentioning

confidence: 99%

Experience and Performance of Persistent Memory for the DUNE Data Acquisition System

Abud

Miotto

Sipos

2021

IEEE Trans. Nucl. Sci.

View full text Add to dashboard Cite

Emerging high-performance storage technologies are opening up the possibility of designing new distributed data acquisition system architectures, in which the live acquisition of data and their processing are decoupled through a storage element. An example of these technologies is 3DXPoint, which promises to fill the gap between memory and traditional storage and offers unprecedented high throughput for non-volatile data.In this paper, we characterize the performance of persistent memory devices, which use the 3DXPoint technology, in the context of the data acquisition system for one large Particle Physics experiment, DUNE. This experiment must be capable of storing, upon a specific signal, incoming data for up to 100 seconds, with a throughput of 1.5 TB/s, for an aggregate size of 150 TB. The modular nature of the apparatus allows splitting the problem into 150 identical units operating in parallel, each at 10 GB/s. The target is to be able to dedicate a single CPU to each of those units for data acquisition and storage.

show abstract

“…For the Intel Optane DC Persistent DIMM, the device is integrated into the system with a DIMM-based interface, similar to DRAM devices. The system directly accesses the device using load/store requests at the byte granularity [95,106,[178][179][180][181][182][183][184]. This configuration provides a much lower access latency, on the order of hundreds of nanoseconds (around 169 ns for sequential reads [95,184]), but comes at a high cost, 5× the cost of the Intel Optane SSD in dollars-per-bit [104,185].…”

Section: Intel Optane Ssdmentioning

confidence: 99%