Pith. sign in

REVIEW 4 major objections 7 minor 54 references

From Good to Great: Improving Memory Tiering Performance Through Parameter Tuning

T0 review · 4 major / 7 minor · reviewed 2026-08-16 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Existing memory tiering systems can run up to 2x faster when a Bayesian optimizer tunes their configuration knobs.

desk verdict Useful first application of Bayesian optimization to memory-tiering knob tuning, with honest transfer experiments, but the headline 2x claim rests on a modified HeMem baseline and needs statistical and artifact backing. read the letter →

arxiv 2504.18714 v1 pith:XIFRXEWP submitted 2025-04-25 cs.OS

Konstantinos Kanellis , Sujay Yadalam , Fanchao Chen , Michael Swift , Shivaram Venkataraman This is my paper · ORCID
classification cs.OS
keywords memorytieringBayesianoptimizationparametertuningpagemigrationhotandcoldpagesHeMemHMSDKCXL
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper argues that existing memory tiering systems leave performance on the table because they rely on static, pre-configured thresholds that do not adapt to workload behavior or hardware. It proposes treating a tiering system as a black box and using Bayesian optimization to search its parameter space, finding that a handful of tuned knob values can improve performance by up to 2x over default configurations and 1.56x over Memtis, a state-of-the-art dynamic system. The paper also shows that the best settings depend on workload, input, thread count, and fast-to-slow memory ratio, and that hidden implementation parameters such as HeMem's cooling_pages can matter as much as the documented ones. If true, this suggests that tiering systems should be designed to adapt all their parameters, not just one or two thresholds.

What carries the argument

The central object is the configuration space of a tiering system's parameters, searched by Sequential Model-based Algorithm Configuration (SMAC), a Bayesian optimization framework whose random-forest surrogate predicts execution time for candidate knob settings and whose acquisition function balances exploration and exploitation. The argument is carried by HeMem's knobs, including read_hot_threshold, write_hot_threshold, cooling_threshold, sampling_period, migration_period, and the hidden cooling_pages, plus a newly exposed write_sampling_period; the optimizer recompiles HeMem with each suggested setting. The same pipeline tunes HMSDK's DAMON-based monitoring parameters. The key mechanism is that different knob combinations produce different page migration patterns, so the optimizer's search over knobs effectively searches over data-placement policies.

What would settle it

Run the published HeMem code unmodified on the same eight workloads with the paper's best-per-configuration knob values; if the tuned values do not beat the stock default on most workloads, the claim that tuning existing systems yields up to 2x is contingent on the authors' modifications. A quicker check: measure whether the stock HeMem default, with only the write_sampling_period knob added, already accounts for most of the performance gap.

Watch

Extended reading notes

Core claim

On its own terms, the paper establishes that the parameters (knobs) of existing tiering systems like HeMem and HMSDK encode enough flexibility to realize workload-specific data placement and migration policies. A Bayesian optimizer (SMAC) given a budget of 100 runs learns which knob values minimize execution time for a given workload and hardware, converging sample-efficiently. Across eight diverse workloads, the best found configuration outperforms the default by 1.07–2.09x; the strongest cases come from workloads with streaming access patterns where the default wastes bandwidth on migrations. The paper also demonstrates that tuning benefits generalize across machines, inputs, thread counts, and memory ratios, though gains shrink when tiers have similar bandwidth (NUMA-emulated CXL).

Load-bearing premise

The headline 2x improvement is measured on a modified version of HeMem that includes four bug fixes and a new write-sampling knob; if those changes, rather than the tuned parameter values, are what separate the default from the tuned configuration, the result does not transfer to the publicly available HeMem.

Editorial extensions

If this is right

  • Existing tiering systems can be improved without redesigning their policies: a search over their existing knobs yields 1.07–2.09x speedups over their default configurations on the workloads tested.
  • The best parameter values are not portable: they change with workload, input dataset, thread count, and fast-to-slow memory ratio, so a single global configuration cannot capture the gains.
  • Hidden or undocumented parameters, such as HeMem's cooling_pages, can be as important as documented thresholds, so tuning should expose and search the full implementation parameter space.
  • Dynamically adapting only one threshold, as Memtis does, is insufficient; the evaluation shows a tuned HeMem configuration outperforms Memtis on every workload in the set.
  • On hardware where tiers have similar bandwidth, such as the NUMA-emulated CXL setup, tuning gains are small because migration costs and benefits are nearly balanced.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The paper's single-run, offline tuning could be combined with phase detection: re-running the optimizer at each program phase would likely improve workloads like Btree that mix write-heavy and read-heavy phases, a direction the authors acknowledge they did not pursue.
  • The tuned configurations and their measured access patterns form a dataset on which a cheaper predictor could be trained, letting a deployment skip the expensive search once patterns are recognized.
  • The DAMON limitation for GUPS suggests that region-based access monitoring cannot be fixed by parameter tuning; systems with scattered hot pages need finer-grained sampling, independent of knob values.
  • The cost-benefit view of migrations implies that bandwidth-aware migration throttling, rather than static migration_period, could capture many of the tuning gains automatically.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 7 minor

Summary. This paper proposes using Bayesian Optimization (SMAC) to tune the configuration knobs of existing memory tiering systems (HeMem and HMSDK). It reports a grid-search case study showing that two HeMem thresholds cause large performance differences, then applies BO to select knob values per workload. The evaluation reports up to 2.09x improvement over HeMem's default configuration and 1.56x over Memtis, along with analyses of why specific knob changes help (e.g., eliminating useless migrations, adjusting write sampling). It also tests generalization across inputs, thread counts, memory ratios, and machines, and reports a transfer experiment showing that configurations often do not transfer across inputs.

Significance. If the quantitative claims hold, the paper demonstrates a simple, practical mechanism for improving tiered-memory performance without redesigning the tiering engine, and it provides a useful negative result about the limits of configuration transfer. The per-workload analysis of migration patterns is a strength and gives concrete guidance for future adaptive tiering designs. However, the central quantitative claim rests on a modified HeMem implementation whose behavior differs from the stock system, and the experiments lack repeated runs, so the magnitude of the claimed gains is not yet established. The paper is transparent about the transfer failures, which is a point in its favor, but the headline numbers should be interpreted as upper-bound, per-workload fitted improvements rather than expected production gains.

major comments (4)
  1. [§3.2 (Deployment Issues), §4.2 (Figure 2), Table 5] The four modifications to HeMem described in §3.2 change the system being evaluated. Fix 3 reduces the minimum allocation size from 1 GiB to 128 MiB, so workloads whose largest allocations fall in the 128 MiB–1 GiB range are tiered by the modified HeMem but would be ignored by the stock system; the default baseline in Figure 2 is therefore the modified default, not the stock default. Fix 4 introduces a new knob, WRITE_SAMPLING_PERIOD, that does not exist in stock HeMem, and Table 5 credits that knob for the Btree speedup. Consequently, the headline "2x over default" claim (Abstract, §7) has not been established for the publicly available HeMem. Please either repeat the key comparisons on stock HeMem, or scope the claim explicitly and provide an ablation showing the contribution of each modification to the default-vs-best gap.
  2. [§4.1 (Experimental Setup), Figures 2, 6, 7, 9, 10, 11, 13] The reported speedups appear to come from a single evaluation per configuration. The best configuration is selected by BO as the minimum of a noisy objective, and the improvement is measured against one run of the default; without repeated runs or confidence intervals, the magnitude of the improvement (e.g., 2.09x in Figure 2) can be inflated by measurement noise. Please run the default and the final best configuration multiple times (at least 3–5) for the main workloads and report mean and standard deviation, and also report the BO runs (e.g., seeds or multiple optimizer trials) to show the stability of the found configuration.
  3. [§4.6 (Figure 13), Abstract] The comparison to Memtis is asymmetric. HeMem-Best is a per-workload configuration selected by the optimizer after 100 iterations of evaluating that same workload, whereas Memtis is evaluated only with its default configuration and, as far as the paper states, without any tuning of its parameters. The "1.56x over state-of-the-art tiering system" claim in the Abstract thus measures tuned HeMem against untuned Memtis, not parameter tuning generally. Either tune Memtis's own knobs with the same BO pipeline and report that result, or explicitly narrow the claim to "tuned HeMem beats untuned Memtis" and discuss whether Memtis would also benefit from tuning.
  4. [§4.3 (Figure 7), §7 (Conclusion)] The transfer experiments in §4.3 show that the best configuration for one input often performs worse than the default on another input (Figure 7). This is an honest and valuable result, but it also implies that the "up to 2x" gain is a fitted maximum for a workload whose full access behavior is known in advance, hence not the expected gain in an online or production setting. The conclusion should explicitly state that the reported improvements are upper-bound estimates for the offline per-workload tuning scenario, and should qualify the "2x" headline accordingly.
minor comments (7)
  1. [§2 (Figure 1)] The execution-time heatmaps in Figure 1 do not indicate whether each cell is a single run or aggregated over multiple runs; please clarify.
  2. [§3.2 (item 1)] The claim that removing PERF_SAMPLE_WEIGHT does not affect sampling accuracy is not validated; please provide evidence, such as a comparison of sampled access counts before and after the change.
  3. [Abstract and Introduction] The spelling of "HeMem" is inconsistent (e.g., "Hemem" appears in the Abstract); please standardize throughout.
  4. [Table 3] The far-memory bandwidth column lists two values (e.g., 7.45/2.25) without defining the two measures; please specify whether these are read/write bandwidths.
  5. [§4.4.1 (Figure 9a)] The thread-count labels on the x-axis are not defined in the caption; please clarify what "2 4 6 8" and "12" refer to.
  6. [§4.4.2 (Figure 9b)] The memory-ratio labels ("2-1", "1-1", "1-2", "1-4", "1-8", "1-16") are not explained in the caption; please define the ordering, e.g., fast:slow tier size ratio.
  7. [Throughout] The paper does not state whether the modified HeMem code and tuning pipeline will be released; a reproducibility statement or artifact pointer would strengthen the work.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper's tuning improvements are empirical search results, not predictions derived from their own inputs.

full rationale

The paper's central claim is that Bayesian optimization can find tiering-system knob values that outperform default configurations. This is presented as a measured optimization result, not as a prediction from a fitted model. The optimizer evaluates actual workload executions and selects configurations by observed execution time, so the reported 2x improvement is an empirical outcome rather than a claim derived from assumptions. Critically, the paper does not rest on an in-sample prediction: Section 4.3 explicitly tests transferring the best configuration to a different input and reports that it often performs worse than default, and Section 4.4.3 similarly tests cross-machine transfer. This acknowledgment of failed transfer shows the authors are not relabeling a fitted point as a general prediction. The grid-search case study in Section 2 independently demonstrates that non-default configurations improve performance, supporting the existence claim without relying on the optimizer. Self-citations to LlamaTune and TUNA are methodological background for Bayesian optimization and are not load-bearing for the memory-tiering result. The modifications to HeMem described in Section 3.2, including the added WRITE_SAMPLING_PERIOD knob, raise a legitimate external-validity question about whether the headline applies to stock HeMem, but this is a fairness and reproducibility concern, not a circularity: the paper does not define the result in terms of the modified system in a way that forces the conclusion. No load-bearing step reduces to its own inputs by construction.

Assumptions & free parameters 12 free parameters · 4 assumptions · 0 invented entities

The central claim rests on the tunability of HeMem and HMSDK knobs, on the stability of execution time measurements, and on the representativeness of the modified HeMem. The tuned knob values themselves are fitted variables rather than fixed assumptions; they are listed as free parameters. No new physical entities are introduced.

free parameters (12)
  • read_hot_threshold
    Minimum read-access samples to classify a page hot in HeMem. BO tunes it per workload, e.g., 4-5 for GapBS-BC and high values for GUPS, but exact values are not tabulated.
  • write_hot_threshold
    Controls write-hot classification. BO increases it for Btree to avoid write-heavy migrations during initialization.
  • cooling_threshold
    Controls when page access counts are halved. BO tunes it to avoid cooling-induced demotions of actually hot pages.
  • sampling_period
    PEBS sampling interval for loads. BO lowers it for GUPS to improve sampling accuracy.
  • write_sampling_period
    New knob introduced by the authors to control store sampling. Tuned to deprioritize writes in read-only workloads.
  • migration_period
    Migration thread interval. BO tunes it for timely migration, e.g., lowest value for TPC-C.
  • max_migration_rate
    Maximum migration bandwidth cap. Part of the 10-knob search space, though not discussed in detail.
  • cooling_pages
    Number of pages cooled at a time. BO increases it for Silo to cool all pages together, reducing warm-page migrations.
  • hot_ring_reqs_threshold
    Number of hot pages processed at a time. Part of the search space, not individually analyzed.
  • cold_ring_reqs_threshold
    Number of cold pages processed at a time. Part of the search space, not individually analyzed.
  • HMSDK DAMON nr_regions
    Number of DAMON regions. BO increases it for GapBS-PR and Btree on the NUMA machine to improve hot region detection.
  • HMSDK DAMON sample_us
    DAMON sampling interval. BO increases scanning frequency for some workloads.
assumptions (4)
  • domain assumption Execution time measurements are stable enough that single runs suffice to compare configurations.
    The paper reports no error bars or repeated runs in Section 4, so all improvement numbers assume low run-to-run variance.
  • ad hoc to paper The modified HeMem (bug fixes plus the new write_sampling_period knob) remains representative of the original HeMem's tiering behavior.
    Section 3.2 lists four code changes; the default-vs-tuned comparison is performed on this modified system, not on the stock release.
  • domain assumption The selected workloads and thread counts saturate memory bandwidth, making tiering decisions performance-critical.
    Section 4.1 says thread counts are chosen to saturate bandwidth, but no measurement is shown to confirm saturation.
  • domain assumption The knob ranges provided to BO (Table 2) include the practically useful region of the configuration space.
    If the true optimum lies outside these ranges, the reported gains would understate or miss the best configuration.

how reviews work

0 comments
Cite this review

Pith. "Pith review of From Good to Great: Improving Memory Tiering Performance Through Parameter Tuning." pith.science (2026). https://pith.science/paper/XIFRXEWP

@misc{pith2026250418714,
  author       = {Pith},
  title        = {Pith review of: From Good to Great: Improving Memory Tiering Performance Through Parameter Tuning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XIFRXEWP}},
  note         = {Machine review of arXiv:2504.18714}
}
read the original abstract

Memory tiering systems achieve memory scaling by adding multiple tiers of memory wherein different tiers have different access latencies and bandwidth. For maximum performance, frequently accessed (hot) data must be placed close to the host in faster tiers and infrequently accessed (cold) data can be placed in farther slower memory tiers. Existing tiering solutions employ heuristics and pre-configured thresholds to make data placement and migration decisions. Unfortunately, these systems fail to adapt to different workloads and the underlying hardware, so perform sub-optimally. In this paper, we improve performance of memory tiering by using application behavior knowledge to set various parameters (knobs) in existing tiering systems. To do so, we leverage Bayesian Optimization to discover the good performing configurations that capture the application behavior and the underlying hardware characteristics. We find that Bayesian Optimization is able to learn workload behaviors and set the parameter values that result in good performance. We evaluate this approach with existing tiering systems, HeMem and HMSDK. Our evaluation reveals that configuring the parameter values correctly can improve performance by 2x over the same systems with default configurations and 1.56x over state-of-the-art tiering system.

Figures

Figures reproduced from arXiv: 2504.18714 by the authors.

Figure 1
Figure 1. Execution time (in seconds) of GUPS (left) and Silo (right) workloads, when we tweak two HeMem parameters. Default configuration execution time is shown in red box. such as hotness thresholds and migration period which con￾trol the system behavior (see table 2). Developers of these systems perform limited sensitivity studies to find the best values for these parameters which we refer to as the de￾fault values. We co… view at source ↗
Figure 3
Figure 3. GapBS-BC: Num￾ber of migrations over time [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 5
Figure 5. XSBench: Memory access pattern over time. GapBS-BC GapBS-PR Silo GUPS 0 1 2 Normalized Performance 1.25x 1.02x 1.59x 2.03x Default Config Best Config [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
Figures from the paper (3 more)
Figure 8
Figure 8. Figure 8: GapBS-BC memory access pattern over time, for Kronecker-s29 (left) and Twitter (right) input graphs. the other input. For example, we ran GapBS-BC on a twitter graph using the best configuration obtained for the kro￾necker graph [PITH_FULL_IMAGE:figures/full_fig_p009_8.png]
Figure 10
Figure 10. Figure 10: Performance gains (over default) of HeMem’s best configuration for NUMA machine (annotated). We also show how pmem-large best config performed when deployed on NUMA machine. sizes of the fast and slow tier are configured based on the measured RSS of the workload. Figu…
Figure 12
Figure 12. Figure 12: Heatmap of GUPS generated by DAMON, using default HMSDK scanning frequency (left) and high scanning frequency (right) a large number of unnecessary migrations (about 10 million pages). The optimizer identifies knob settings that elimi￾nate migrations of such pages wit…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

54 extracted references · 46 canonical work pages

  1. [1]

    Reto Achermann and Ashish Panwar. 2020. Mitosis Workload Btree. https://github.com/mitosis-project/mitosis-workload-btree

    work page 2020

  2. [2]

    Shashank Adavally and Krishna Kavi. 2021. Subpage migration in het- erogeneous memory systems. In Workshop on Heterogeneous Memory Systems (HMEM-2021), Colocated with ICS 2021

    work page 2021

  3. [3]

    Neha Agarwal and Thomas F Wenisch. 2017. Thermostat: Application- transparent page management for two-tiered main memory. In Pro- ceedings of the Twenty-Second International Conference on Architectural Support for Programming Languages and Operating Systems . 631–644

    work page 2017

  4. [4]

    Scott Beamer, Krste Asanović, and David Patterson. 2015. The GAP benchmark suite. arXiv preprint arXiv:1508.03619 (2015)

    arXiv 2015

  5. [5]

    Zhen Cao, Geoff Kuenning, and Erez Zadok. 2020. Carver: Finding Important Parameters for Storage System Tuning. In18th USENIX Con- ference on File and Storage Technologies (FAST 20). USENIX Association, Santa Clara, CA, 43–57. https://www.usenix.org/conference/fast20/ presentation/cao-zhen

    work page 2020

  6. [6]

    Zhen Cao, Vasily Tarasov, Sachin Tiwari, and Erez Zadok. 2018. To- wards Better Understanding of Black-box{Auto-Tuning}: A Compar- ative Analysis for Storage Systems. In 2018 USENIX Annual Technical Conference (USENIX ATC 18). 893–907

    work page 2018

  7. [7]

    Intel Corporation. 2018. Intel 64 and IA-32 Architectures Software De- veloper Manuals. https://software.intel.com/articles/intel-sdm. (2018)

    work page 2018

  8. [8]

    Ian Cutress and Billy Tallis. 2018. Intel Launches Optane DIMMs up to 512GB: Apache Pass is Here.https://www.anandtech.com/show/12828/ intel-launches-optane-dimms-up-to-512gb-apache-pass-is-here

    work page 2018

Show all 54 references
  1. [9]

    Thaleia Dimitra Doudali, Sergey Blagodurov, Abhinav Vishnu, Sud- hanva Gurumurthi, and Ada Gavrilovska. 2019. Kleio: A hybrid mem- ory page scheduler with machine intelligence. InProceedings of the 28th International Symposium on High-Performance Parallel and Distributed Compu...

    2019

  2. [10]

    Thaleia Dimitra Doudali, Daniel Zahka, and Ada Gavrilovska. 2021. Cori: Dancing to the right beat of periodic data movements over hybrid memory systems. In 2021 IEEE International Parallel and Distributed Processing Symposium (IPDPS). IEEE, 350–359

    2021

  3. [11]

    Subramanya R Dulloor, Amitabha Roy, Zheguang Zhao, Narayanan Sundaram, Nadathur Satish, Rajesh Sankaran, Jeff Jackson, and Karsten Schwan. 2016. Data tiering in heterogeneous memory systems. In Proceedings of the Eleventh European Conference on Computer Systems . 1–16

    2016

  4. [12]

    Dmitry Duplyakin, Robert Ricci, Aleksander Maricq, Gary Wong, Jonathon Duerig, Eric Eide, Leigh Stoller, Mike Hibler, David John- son, Kirk Webb, Aditya Akella, Kuangching Wang, Glenn Ricart, Larry Landweber, Chip Elliott, Michael Zink, Emmanuel Cecchet, Snigdhaswin Kar, and P...

    2019

  5. [13]

    Padmapriya Duraisamy, Wei Xu, Scott Hare, Ravi Rajwar, David Culler, Zhiyi Xu, Jianing Fan, Christopher Kennelly, Bill McCloskey, Danijela Mijailovic, et al. 2023. Towards an adaptable systems architecture for memory tiering at warehouse-scale. In Proceedings of the 28th ACM I...

    2023

  6. [14]

    Ayat Fekry, Lucian Carata, Thomas Pasquier, Andrew Rice, and Andy Hopper. 2020. To Tune or Not to Tune? In Search of Optimal Config- urations for Data Analytics. In Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining (Vir- tual Even...

    2020 doi

  7. [15]

    Johannes Freischuetz, Konstantinos Kanellis, Brian Kroth, and Shiv- aram Venkataraman. 2025. TUNA: Tuning Unstable and Noisy Cloud Applications. In Proceedings of the Twentieth European Conference on Computer Systems (Rotterdam, Netherlands) (EuroSys ’25) . As- sociation for C...

    2025

  8. [16]

    Zhiyuan Guo, Zijian He, and Yiying Zhang. 2023. Mira: A program- behavior-guided far memory system. In Proceedings of the 29th Sym- posium on Operating Systems Principles . 692–708

    2023

  9. [17]

    Taekyung Heo, Yang Wang, Wei Cui, Jaehyuk Huh, and Lintao Zhang

  10. [18]

    Frank Hutter, Holger H Hoos, and Kevin Leyton-Brown. 2011. Sequen- tial model-based optimization for general algorithm configuration. In Learning and Intelligent Optimization: 5th International Conference, LION 5, Rome, Italy, January 17-21, 2011. Selected Papers 5 . Springer, 507–523

    2011

  11. [19]

    Compute Express Link Consortium. Inc. 2020. CXL ® Specification. https://computeexpresslink.org/cxl-specification/

    2020

  12. [20]

    Joseph Izraelevitz, Jian Yang, Lu Zhang, Juno Kim, Xiao Liu, Amir- saman Memaripour, Yun Joon Soh, Zixuan Wang, Yi Xu, Subramanya R Dulloor, et al. 2019. Basic performance measurements of the intel op- tane DC persistent memory module. arXiv preprint arXiv:1903.05714 (2019)

    2019 arXiv

  13. [21]

    Konstantinos Kanellis, Ramnatthan Alagappan, and Shivaram Venkataraman. 2020. Too many knobs to tune? towards faster database tuning by pre-selecting important knobs. In 12th USENIX Workshop on Hot Topics in Storage and File Systems (HotStorage 20)

    2020

  14. [22]

    Konstantinos Kanellis, Cong Ding, Brian Kroth, Andreas Müller, Carlo Curino, and Shivaram Venkataraman. 2022. LlamaTune: Sample- efficient DBMS configuration tuning. arXiv preprint arXiv:2203.05128 (2022)

    2022 arXiv

  15. [23]

    KyungSoo Lee, Sohyun Kim, Joohee Lee, Donguk Moon, Rakie Kim, Honggyu Kim, Hyeongtak Ji, Yunjeong Mun, and Youngpyo Joo. 2024. Improving key-value cache performance with heterogeneous memory tiering: A case study of CXL-based memory expansion. IEEE Micro (2024)

    2024

  16. [24]

    Taehyung Lee, Sumit Kumar Monga, Changwoo Min, and Young Ik Eom. 2023. MEMTIS: Efficient Memory Tiering with Dynamic Page Classification and Page Size Determination. In Proceedings of the 29th Symposium on Operating Systems Principles . 17–34. 13

    2023

  17. [25]

    Baptiste Lepers and Willy Zwaenepoel. 2023. Johnny Cache: the End of {DRAM} Cache Conflicts (in Tiered Main Memory Systems). In 17th USENIX Symposium on Operating Systems Design and Implementation (OSDI 23). 519–534

    2023

  18. [26]

    2022.{CBMM}: Financial Advice for Kernel Memory Managers

    Mark Mansi, Bijan Tabatabai, and Michael M Swift. 2022.{CBMM}: Financial Advice for Kernel Memory Managers. In2022 USENIX Annual Technical Conference (USENIX ATC 22). 593–608

    2022

  19. [27]

    Adnan Maruf, Ashikee Ghosh, Janki Bhimani, Daniel Campello, Andy Rudoff, and Raju Rangaswami. 2022. MULTI-CLOCK: Dynamic Tiering for Hybrid Memory Systems.. In HPCA. 925–937

    2022

  20. [28]

    Hasan Al Maruf, Hao Wang, Abhishek Dhanotia, Johannes Weiner, Niket Agarwal, Pallab Bhattacharya, Chris Petersen, Mosharaf Chowd- hury, Shobhit Kanaujia, and Prakash Chauhan. 2023. TPP: Transparent page placement for CXL-enabled tiered-memory. In Proceedings of the 28th ACM In...

    2023

  21. [29]

    Dingheng Mo, Fanchao Chen, Siqiang Luo, and Caihua Shan. 2023. Learning to Optimize LSM-trees: Towards A Reinforcement Learning based Key-Value Store for Dynamic Workloads.Proc. ACM Manag. Data 1, 3, Article 213 (Nov. 2023), 25 pages. https://doi.org/10.1145/3617333

    2023 doi

  22. [30]

    Timothy Morgan. 2020. CXL And Gen-Z Iron Out A Coherent In- terconnect Strategy. https://www.nextplatform.com/2020/04/03/cxl- and-gen-z-iron-out-a-coherent-interconnect-strategy/ . In The Next Platform

    2020

  23. [31]

    Richard C Murphy, Kyle B Wheeler, Brian W Barrett, and James A Ang

  24. [32]

    SeongJae Park, Yunjae Lee, and Heon Y Yeom. 2019. Profiling dynamic data access patterns with controlled overhead and quality. In Proceed- ings of the 20th International Middleware Conference Industrial Track . 1–7

    2019

  25. [33]

    Steven J Plimpton, Ron Brightwell, Courtenay Vaughan, Keith Un- derwood, and Mike Davis. 2006. A simple synchronous distributed- memory algorithm for the HPCC RandomAccess benchmark. In 2006 IEEE International Conference on Cluster Computing . IEEE, 1–7

    2006

  26. [34]

    Luiz E Ramos, Eugene Gorbatov, and Ricardo Bianchini. 2011. Page placement in hybrid memory systems. In Proceedings of the interna- tional conference on Supercomputing . 85–95

    2011

  27. [35]

    Amanda Raybuck, Tim Stamler, Wei Zhang, Mattan Erez, and Simon Peter. 2021. Hemem: Scalable tiered memory management for big data applications and real nvm. In Proceedings of the ACM SIGOPS 28th Symposium on Operating Systems Principles . 392–407

    2021

  28. [36]

    Jee Ho Ryoo, Lizy K John, and Arkaprava Basu. 2018. A case for gran- ularity aware page migration. In Proceedings of the 2018 International Conference on Supercomputing. 352–362

    2018

  29. [37]

    Bobak Shahriari, Kevin Swersky, Ziyu Wang, Ryan P Adams, and Nando De Freitas. 2015. Taking the human out of the loop: A review of Bayesian optimization. Proc. IEEE 104, 1 (2015), 148–175

    2015

  30. [38]

    Gagandeep Singh, Rakesh Nadig, Jisung Park, Rahul Bera, Nastaran Hajinazar, David Novo, Juan Gómez-Luna, Sander Stuijk, Henk Cor- poraal, and Onur Mutlu. 2022. Sibyl: Adaptive and extensible data placement in hybrid storage systems using online reinforcement learn- ing. In Pro...

    2022

  31. [39]

    Jasper Snoek, Hugo Larochelle, and Ryan P Adams. 2012. Practical bayesian optimization of machine learning algorithms. Advances in Neural Information Processing Systems 25 (2012)

    2012

  32. [40]

    Yan Sun, Yifan Yuan, Zeduo Yu, Reese Kuper, Chihun Song, Jinghan Huang, Houxiang Ji, Siddharth Agarwal, Jiaqi Lou, Ipoom Jeong, et al

  33. [41]

    John R Tramm, Andrew R Siegel, Tanzima Islam, and Martin Schulz

  34. [42]

    Stephen Tu, Wenting Zheng, Eddie Kohler, Barbara Liskov, and Samuel Madden. 2013. Speedy transactions in multicore in-memory databases. In Proceedings of the Twenty-Fourth ACM Symposium on Operating Systems Principles. 18–32

    2013

  35. [43]

    Johannes Weiner, Niket Agarwal, Dan Schatzberg, Leon Yang, Hao Wang, Blaise Sanouillet, Bikash Sharma, Tejun Heo, Mayank Jain, Chunqiang Tang, et al. 2022. TMO: Transparent memory offloading in datacenters. In Proceedings of the 27th ACM International Conference on Architectur...

    2022

  36. [44]

    Lingfeng Xiang, Zhen Lin, Weishu Deng, Hui Lu, Jia Rao, Yifan Yuan, and Ren Wang. 2024. MATRYOSHKA: Non-Exclusive Memory Tiering via Transactional Page Migration. arXiv preprint arXiv:2401.13154 (2024)

    2024 arXiv

  37. [45]

    Qiumin Xu, Huzefa Siyamwala, Mrinmoy Ghosh, Tameesh Suri, Manu Awasthi, Zvika Guz, Anahita Shayesteh, and Vijay Balakrishnan. 2015. Performance analysis of NVMe SSDs and their implication on real world databases. In Proceedings of the 8th ACM International Systems and Storage ...

    2015

  38. [46]

    Zi Yan, Daniel Lustig, David Nellans, and Abhishek Bhattacharjee. 2019. Nimble page management for tiered memory systems. In Proceedings of the Twenty-Fourth International Conference on Architectural Support for Programming Languages and Operating Systems . 331–345

    2019

  39. [47]

    Zhengyu Yang, Morteza Hoseinzadeh, Allen Andrews, Clay Mayers, David Thomas Evans, Rory Thomas Bolt, Janki Bhimani, Ningfang Mi, and Steven Swanson. 2017. AutoTiering: Automatic data place- ment manager in multi-tier all-flash datacenter. In 2017 IEEE 36th International Perfor...

    2017

  40. [48]

    Seongdae Yu, Seongbeom Park, and Woongki Baek. 2017. Design and implementation of bandwidth-aware memory placement and migra- tion policies for heterogeneous memory systems. In Proceedings of the International Conference on Supercomputing . 1–10

    2017

  41. [49]

    Lei Zhang, Reza Karimi, Irfan Ahmad, and Ymir Vigfusson. 2020. Op- timal data placement for heterogeneous cache, memory, and storage systems. Proceedings of the ACM on Measurement and Analysis of Computing Systems 4, 1 (2020), 1–27

    2020

  42. [50]

    Chenxingyu Zhao, Tapan Chugh, Jaehong Min, Ming Liu, and Arvind Krishnamurthy. 2022. Dremel: Adaptive Configuration Tuning of RocksDB KV-Store. Proc. ACM Meas. Anal. Comput. Syst. 6, 2, Article 37 (June 2022), 30 pages. https://doi.org/10.1145/3530903 14

    2022 doi

  43. [2010]

    Cray Users Group (CUG) 19, 45-74 (2010), 22

    Introducing the graph 500. Cray Users Group (CUG) 19, 45-74 (2010), 22

    2010

  44. [2014]

    The Role of Reactor Physics toward a Sustainable Future (PHYSOR) (2014)

    XSBench-the development and verification of a performance abstraction for Monte Carlo reactor analysis. The Role of Reactor Physics toward a Sustainable Future (PHYSOR) (2014)

    2014

  45. [2020]

    IEEE Trans

    Adaptive page migration policy with huge pages in tiered mem- ory systems. IEEE Trans. Comput. 71, 1 (2020), 53–68

    2020

  46. [2023]

    In Proceedings of the 56th Annual IEEE/ACM International Symposium on Microarchitecture

    Demystifying cxl memory with genuine cxl-ready systems and devices. In Proceedings of the 56th Annual IEEE/ACM International Symposium on Microarchitecture. 105–121

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.