pith. sign in

arxiv: 2605.20906 · v1 · pith:LIICHA57new · submitted 2026-05-20 · 💻 cs.OS

ParaCell: Paravirtualized Secure Containers with Lightweight Intra-Container Isolation and Intent-Driven Memory Management

Yiyang Wu , Xunjie Wang , Jinyu Gu , Haibo Chen This is my paper

Pith reviewed 2026-05-21 02:09 UTC · model grok-4.3

classification 💻 cs.OS
keywords secure containersparavirtualizationMPK isolationmemory managementintra-container isolationcloud performanceagent workloadspage binding
0
0 comments X

The pith

ParaCell uses MPK domain switches and proactive memory binding to cut secure container latency while preserving elasticity.

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

The paper presents ParaCell, a paravirtualized secure container runtime that isolates each container with its own kernel. It addresses the isolation-performance trade-off by applying hardware memory protection keys to isolate user and kernel code inside one address space, turning transitions into fast domain switches instead of VM exits. It also reads memory-management intent directly from container kernel allocators to batch proactive guest-to-host page bindings and avoid reactive faults. Experiments across cloud and agent workloads report large latency drops versus PVM and RunV plus memory savings versus HyperAlloc. If correct, this would let secure containers handle both traditional services and bursty agent tasks more efficiently without new hardware or coarser memory granularity.

Core claim

ParaCell achieves its results by combining two mechanisms inside a drop-in replacement for RunV: MPK-based XGates that isolate the container user and container kernel within a single address space so that user-kernel transitions become direct domain switches, and a Pager component that interposes on allocation and free events to perform batch proactive GPA-to-HPA bindings and unbindings, eliminating most reactive shadow page-table faults while keeping fine-grained elasticity.

What carries the argument

MPK-based XGates for intra-address-space user-kernel isolation plus the Pager that extracts memory intent from container kernel allocators to drive proactive page bindings.

If this is right

  • Latency drops of up to 57 percent and 79 percent versus PVM in bare-metal and nested setups.
  • Latency drops of up to 33 percent and 88 percent versus RunV in the same setups.
  • Memory savings up to 35.6 percent versus HyperAlloc on agent workloads.
  • Fine-grained memory elasticity is retained while avoiding most secondary page faults.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same intent-extraction idea could be applied to other guest kernels or hypervisor interfaces where allocation events are visible.
  • If MPK-style primitives become more common in future CPUs, similar lightweight isolation layers may appear in additional virtualization stacks.
  • Agent frameworks that already expose memory hints might integrate directly with the Pager to further improve elasticity.

Load-bearing premise

MPK protection primitives can deliver lightweight isolation for frequent user-kernel transitions without new security risks or hidden costs, and container kernel allocators already encode enough memory-management intent to make proactive GPA-HPA bindings effective.

What would settle it

A measurement showing either that MPK domain switches introduce unacceptable overhead or security holes in container workloads, or that proactive bindings fail to reduce faults below reactive baselines on bursty agent memory patterns.

Figures

Figures reproduced from arXiv: 2605.20906 by Haibo Chen, Jinyu Gu, Xunjie Wang, Yiyang Wu.

Figure 1
Figure 1. Figure 1: Memory characteristics of SWE-bench agent exe￾cution. (a) In-use guest memory at different page granulari￾ties over execution time. (b) Distribution of execution time fraction across memory waste levels. across all sampled SWE-bench executions, memory waste, defined as the extra memory consumed by 2MB pages rela￾tive to 4KB pages, has a mean of 23.3% and a median of 17.1%. Consequently, these patterns dema… view at source ↗
Figure 2
Figure 2. Figure 2: Additional shadow faults incurred by shadow pag￾ing during anonymous page fault handling. Page fault for￾warding ( 1 ) and emulation of guest page table writes ( 7 ) are omitted for brevity. PF: page fault. SF: shadow fault. GFP: get_free_pages. U-PT: page table of the faulting process. U-SPT: shadow page table of the faulting process. DM-SPT: shadow page table of kernel direct mapping. Specifically, it fi… view at source ↗
Figure 3
Figure 3. Figure 3: , ParaCell preserves the VM-like secure container abstraction of a separate guest kernel, while deprivileging the guest kernel to user mode to avoid rigid dependence on hardware-assisted virtualization, following prior PV se￾cure containers [42, 65]. Within each container, ParaCell preserves isolation among multiple user applications and be￾tween user applications and the guest kernel, while enabling effic… view at source ↗
Figure 4
Figure 4. Figure 4: ParaCell’s control-flow-aware domain-switch ar￾chitecture compared with traditional PV switch gates design. U/K: user/kernel. AS: address space. MPK: Memory Protec￾tion Keys. Iso: isolation. GK. Third, external interrupts require host-side acknowledg￾ment and routing, so they must remain on the conventional host-mediated path. Based on this classification, each container owns a pair of XGates rather than a… view at source ↗
Figure 5
Figure 5. Figure 5: Runtime Syscall Deprivileging inside the Syscall Gate. GK: Guest Kernel. GU: Guest User. para_cli/sti: em￾ulated cli/sti with vCPU’s interrupt flag manipulation. Inside the gate, to_kernel handles the forward transition. It saves user execution context on the stack, switches to the GK memory domain, and issues an emulated cli to block in￾terrupt delivery in the critical section. It then restores guest￾kern… view at source ↗
Figure 6
Figure 6. Figure 6: Pager architecture and workflow. PT: page ta￾ble. DM-PT: page table managing direct mapping. GFP: get_free_pages. AS Iso: address space isolation. U/K Iso: user/kernel privilege isolation. On allocation (green in the figure), the Pager allocates a backing page from the host ( 1 4 , as explained below), records the GPA→HPA binding ( 2 ). It also immediately maps the translated HPA into the direct-mapping PT… view at source ↗
Figure 7
Figure 7. Figure 7: System benchmark latency on LMbench. 2M: 2MB huge-page mapping for VM (container) memory. NST: nested virtualization. 275ns (7%), thanks to the lightweight switch overhead to Pager. The batched GPA→HPA binding (plus HPA alloca￾tion) is amortized to 175ns (4%) per page. Finally, XGates reduces the GU/GK switch latency to 1622ns, approaching the RunV result of 1028ns. ParaCell reduces the fault latency of co… view at source ↗
Figure 9
Figure 9. Figure 9: Normalized latency on sqlite-bench. 2M: 2MB huge-page mapping for VM (container) memory. NST: nested virtualization. from lightweight intra-container isolation. For larger batched operations, where page fault overhead becomes more visible, ParaCell reduces latency by up to 33.5% and 14.9% compared with PVM and RunV, respectively. On the largest batch (de￾noted by the 100K suffix), ParaCell reduces latency … view at source ↗
Figure 10
Figure 10. Figure 10: Normalized latency on memory-intensive work￾loads. 2M: 2MB huge-page mapping for VM (container) mem￾ory. NST: nested virtualization. Page-fault-intensive workloads. As shown in [PITH_FULL_IMAGE:figures/full_fig_p011_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Normalized end-to-end latency on SkillsBench. 2M: 2MB huge-page mapping for VM (container) memory. NST: nested virtualization. to the guest’s truly in-use 4KB memory. On SkillsBench, ParaCell’s host-allocated memory closely tracks guest in-use memory throughout the execution, resulting in only 0.2% mean memory overhead. In contrast, HyperAlloc suffers from internal fragmentation within huge pages, leading… view at source ↗
Figure 12
Figure 12. Figure 12: Distribution of execution time fraction across host-allocated memory overhead levels during SWE-bench runs, shown at two x-axis scales to expose both the low (left) and high (right) overhead regions. 7 Other Related Work Intra-address-space isolation. Intra-address-space isola￾tion is a technique to achieve lightweight isolation across different components within a single address space. Various systems le… view at source ↗
read the original abstract

Secure containers isolate each container with its own kernel, mitigating shared-kernel attacks prevalent in traditional container systems. However, existing designs still face a fundamental isolation--performance trade-off. Nested-cloud deployments amplify the cost of VM exits and page-table management, while emerging agentic workloads expose bursty memory demand that requires fine-grained elasticity. We attribute this trade-off to two root causes. First, existing designs lack lightweight intra-container isolation primitives for frequent container user--kernel transitions. Second, the host treats container memory management as opaque, forcing reactive secondary faults and coarse-grained huge page mappings to amortize their cost. This paper presents ParaCell, a paravirtualized secure container runtime built on two insights. First, intra-address-space hardware protection primitives can provide lightweight intra-container isolation. ParaCell uses MPK-based XGates to isolate the container user and container kernel within a single address space, turning frequent user--kernel transitions into direct domain switches. Second, container kernel allocators already encode memory-management intent. ParaCell introduces Pager to interpose on allocation and free events, batch proactive GPA to HPA bindings and unbindings, and avoid reactive shadow page-table faults while preserving fine-grained memory elasticity. ParaCell is implemented as a drop-in replacement for RunV. Our experiments demonstrate that, across traditional cloud and emerging agent applications, ParaCell reduces latency by up to 57% and 79% over PVM, and by up to 33% and 88% over RunV, in bare-metal and nested setups, respectively. On agent workloads, ParaCell saves up to 35.6% memory compared with the state-of-the-art VM memory reclamation technique, HyperAlloc.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 3 minor

Summary. The paper presents ParaCell, a paravirtualized secure container runtime that uses MPK-based XGates to provide lightweight intra-container isolation by placing container user and kernel code in the same address space and converting syscalls to domain switches, along with the Pager mechanism that interposes on kernel allocator events to perform proactive GPA-to-HPA bindings and avoid reactive faults. Experiments claim latency reductions of up to 57% and 79% versus PVM and 33% and 88% versus RunV in bare-metal and nested configurations, plus up to 35.6% memory savings versus HyperAlloc on agent workloads.

Significance. If the reported performance and memory results hold under full scrutiny, the work would offer a practical advance in secure container design by mitigating the isolation-performance trade-off in both traditional and nested-cloud settings while supporting elastic memory for bursty agentic applications. The combination of hardware domain protection with allocator-intent memory management is a concrete contribution that could influence future paravirtualized runtimes.

major comments (2)
  1. [§3.1] §3.1 (XGates design): The claim that MPK-based domain switches deliver lightweight isolation without new security risks rests on the assumption that the container kernel cannot arbitrarily load PKRU or bypass protections while sharing the address space; the manuscript does not detail the required kernel modifications or prove that TLB invalidations remain minimal in nested virtualization, which is load-bearing for attributing the 57–88% latency gains specifically to the isolation primitive rather than Pager batching.
  2. [§5.2] §5.2 (evaluation methodology): The latency and memory numbers are presented without error bars, full baseline configurations, or raw data tables; this prevents verification that the reported improvements (e.g., 35.6% memory savings) are statistically robust and not confounded by workload-specific tuning or measurement artifacts.
minor comments (3)
  1. [Abstract] The abstract and §2 could more clearly distinguish the contributions of XGates versus Pager so readers can assess which component drives the nested-setup gains.
  2. [Figure 4] Figure 4 (memory elasticity plot) uses overlapping lines without distinct markers or a legend inset, reducing readability for the agent-workload comparison.
  3. [§6] A short related-work paragraph contrasting ParaCell with prior MPK uses in unikernels or library OSes would help situate the novelty of the intra-container kernel isolation approach.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed feedback on our manuscript. The comments help clarify important aspects of the XGates design and evaluation presentation. We respond to each major comment below and indicate the revisions we will make.

read point-by-point responses

  1. Referee: [§3.1] §3.1 (XGates design): The claim that MPK-based domain switches deliver lightweight isolation without new security risks rests on the assumption that the container kernel cannot arbitrarily load PKRU or bypass protections while sharing the address space; the manuscript does not detail the required kernel modifications or prove that TLB invalidations remain minimal in nested virtualization, which is load-bearing for attributing the 57–88% latency gains specifically to the isolation primitive rather than Pager batching.

    Authors: We thank the referee for this observation. Section 3.1 describes the paravirtualized kernel modifications that trap and validate all PKRU writes, restricting the container kernel to only authorized XGates for domain switches; untrusted code cannot load arbitrary PKRU values. On TLB behavior in nested virtualization, MPK domain switches operate within a shared address space and do not trigger additional page-table walks or full TLB flushes beyond those of a standard syscall. To better attribute the reported latency reductions to the isolation mechanism itself, we will add an ablation study in the revised §5 that disables Pager batching while retaining XGates, allowing direct comparison of the two contributions. revision: yes

  2. Referee: [§5.2] §5.2 (evaluation methodology): The latency and memory numbers are presented without error bars, full baseline configurations, or raw data tables; this prevents verification that the reported improvements (e.g., 35.6% memory savings) are statistically robust and not confounded by workload-specific tuning or measurement artifacts.

    Authors: We agree that the current evaluation presentation would benefit from greater statistical transparency. In the revised manuscript we will add error bars (standard deviation over 10 runs) to all latency and memory figures in §5.2 and include a new appendix table that fully specifies the configuration parameters of PVM, RunV, and HyperAlloc. We will also add a compact summary table of the key raw measurements. The complete per-run dataset will be released as supplementary material rather than embedded in the paper to keep the main text concise. revision: partial

Circularity Check

0 steps flagged

No circularity in ParaCell design or claims

full rationale

The paper introduces ParaCell as a new paravirtualized secure container runtime using MPK-based XGates for intra-address-space isolation and a Pager for proactive GPA-HPA bindings based on allocator events. All performance claims (latency reductions of 57-88% and memory savings of 35.6%) are presented as outcomes of implementation and direct experimental measurements against baselines like PVM, RunV, and HyperAlloc. No equations, fitted parameters renamed as predictions, self-citations forming load-bearing uniqueness theorems, or ansatzes smuggled via prior work appear in the provided text. The derivation chain consists of engineering insights implemented and benchmarked, remaining self-contained against external benchmarks without reducing to its own inputs by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 2 invented entities

Central claims depend on the effectiveness of newly introduced primitives whose security and performance properties are asserted but not independently evidenced in the abstract.

axioms (2)
  • domain assumption Intra-address-space hardware protection primitives can provide lightweight intra-container isolation.
    First insight underlying XGates; invoked to justify turning user-kernel transitions into direct domain switches.
  • domain assumption Container kernel allocators encode memory-management intent usable for proactive GPA to HPA bindings.
    Second insight underlying Pager; invoked to enable batch proactive bindings and avoid reactive faults.
invented entities (2)
  • XGates no independent evidence
    purpose: Lightweight intra-container isolation using MPK within a single address space.
    New primitive introduced to address frequent user-kernel transitions.
  • Pager no independent evidence
    purpose: Interpose on allocation/free events for batch proactive memory bindings.
    New component to exploit encoded intent for fine-grained elasticity.

pith-pipeline@v0.9.0 · 5851 in / 1474 out tokens · 59622 ms · 2026-05-21T02:09:02.915273+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

87 extracted references · 87 canonical work pages · 3 internal anchors

  1. [1]

    Computer use.https://platform.openai.com/docs/guides/tools- computer-use

    work page

  2. [2]

    The container security platform.https://gvisor.dev/

    work page

  3. [3]

    db_bench_sqlite3.cc.https://github.com/google/leveldb/blob/main/ benchmarks/db_bench_sqlite3.cc

    work page

  4. [4]

    Deploy openclaw in seconds.https://www.alibabacloud.com/en/ campaign/ai-openclaw

    work page

  5. [5]

    Hermes agent.https://hermes-agent.nousresearch.com/

    work page

  6. [6]

    How claude code works.https://code.claude.com/docs/en/how- claude-code-works

    work page

  7. [7]

    Introducing claude sonnet 4.6.https://www.anthropic.com/news/ claude-sonnet-4-6

    work page

  8. [8]

    Introducing site isolation in firefox.https://blog.mozilla.org/security/ 2021/05/18/introducing-site-isolation-in-firefox/

    work page 2021

  9. [9]

    Introducing upgrades to codex.https://openai.com/index/introducing- upgrades-to-codex/

    work page

  10. [10]

    Kata containers.https://katacontainers.io/

    work page

  11. [11]

    Let claude use your computer from the cli.https://code.claude.com/ docs/en/computer-use

    work page

  12. [12]

    sourceforge.net/

    Lmbench - tools for performance analysis.https://lmbench. sourceforge.net/

    work page

  13. [13]

    Multi-process architecture.https://www.chromium.org/developers/ design-documents/multi-process-architecture/

    work page

  14. [14]

    com/act/pro/intl-openclaw

    Openclaw - deploy openclaw in seconds.https://www.tencentcloud. com/act/pro/intl-openclaw

    work page

  15. [15]

    Openclaw overview.https://docs.openclaw.ai/

    work page

  16. [16]

    Playwright mcp.https://playwright.dev/mcp/introduction

    work page

  17. [17]

    com/act/pro/hermesagent

    Quickly deploy hermes agent on the cloud.https://www.tencentcloud. com/act/pro/hermesagent

    work page

  18. [18]

    runc.https://github.com/opencontainers/runc

    work page

  19. [19]

    runv.https://github.com/hyperhq/runv

    work page

  20. [20]

    Sandboxing.https://docs.openclaw.ai/gateway/sandboxing

    work page

  21. [21]

    Site isolation design document.https://www.chromium.org/ developers/design-documents/site-isolation/

    work page

  22. [22]

    Sqlite.https://www.sqlite.org/

    work page

  23. [23]

    The systrap platform.https://github.com/google/gvisor/blob/master/ pkg/sentry/platform/systrap/README.md

    work page

  24. [24]

    Fire- cracker: Lightweight virtualization for serverless applications

    Alexandru Agache, Marc Brooker, Alexandra Iordache, Anthony Liguori, Rolf Neugebauer, Phil Piwonka, and Diana-Maria Popa. Fire- cracker: Lightweight virtualization for serverless applications. In17th USENIX symposium on networked systems design and implementation (NSDI 20), pages 419–434, 2020

    work page 2020

  25. [25]

    Drops: Managing serverless resource pools in mi- crosoft azure functions

    Ahmed Alquraan, Abdelrahman Baba, Rafael Mendes da Silva, Sameh Elnikety, Paul Batum, Yan Chen, Hamid Henry Safi, Seth Fine, and Samer Al-Kiswany. Drops: Managing serverless resource pools in mi- crosoft azure functions. InProceedings of the 21st European Conference on Computer Systems, pages 1281–1297, 2026

    work page 2026

  26. [26]

    The design and implementation of hyperupcalls

    Nadav Amit and Michael Wei. The design and implementation of hyperupcalls. In2018 USENIX Annual Technical Conference (USENIX ATC 18), pages 97–112, 2018

    work page 2018

  27. [27]

    The turtles project: Design and implementation of nested virtualization

    Muli Ben-Yehuda, Michael D Day, Zvi Dubitzky, Michael Factor, Nadav Har’El, Abel Gordon, Anthony Liguori, Orit Wasserman, and Ben-Ami Yassour. The turtles project: Design and implementation of nested virtualization. In9th USENIX Symposium on Operating Systems Design and Implementation (OSDI 10), 2010

    work page 2010

  28. [28]

    Out of hypervisor (ooh): efficient dirty page tracking in userspace using hardware virtualization features

    Stella Bitchebe and Alain Tchana. Out of hypervisor (ooh): efficient dirty page tracking in userspace using hardware virtualization features. InSC22: International Conference for High Performance Computing, Networking, Storage and Analysis, pages 1–14. IEEE, 2022

    work page 2022

  29. [29]

    An analysis of linux scalability to many cores

    Silas Boyd-Wickizer, Austin T Clements, Yandong Mao, Aleksey Pesterev, M Frans Kaashoek, Robert Morris, and Nickolai Zeldovich. An analysis of linux scalability to many cores. In9th USENIX Sym- posium on Operating Systems Design and Implementation (OSDI 10), 2010

    work page 2010

  30. [30]

    Skernel: An elastic and efficient secure container system at scale with a split- kernel architecture

    Xiaohu Chai, Keyang Hu, Jianfeng Tan, Tiwei Bie, Guotao Tan, Tianyu Zhou, Anqi Shen, Dawei Shen, Xinyao Yang, Xin Chen, et al. Skernel: An elastic and efficient secure container system at scale with a split- kernel architecture. InProceedings of the 21st European Conference on Computer Systems, pages 605–623, 2026

    work page 2026

  31. [31]

    Limitations and opportunities of modern hardware isolation mechanisms

    Xiangdong Chen, Zhaofeng Li, Tirth Jain, Vikram Narayanan, and Anton Burtsev. Limitations and opportunities of modern hardware isolation mechanisms. In2024 USENIX Annual Technical Conference (USENIX ATC 24), pages 349–368, 2024

    work page 2024

  32. [32]

    Catalyzer: Sub-millisecond startup for serverless computing with initialization-less booting

    Dong Du, Tianyi Yu, Yubin Xia, Binyu Zang, Guanglu Yan, Cheng- gang Qin, Qixuan Wu, and Haibo Chen. Catalyzer: Sub-millisecond startup for serverless computing with initialization-less booting. In Proceedings of the Twenty-Fifth International Conference on Architec- tural Support for Programming Languages and Operating Systems, pages 467–481, 2020

    work page 2020

  33. [33]

    Mak- ing kernel bypass practical for the cloud with junction

    Joshua Fried, Gohar Irfan Chaudhry, Enrique Saurez, Esha Choukse, Íñigo Goiri, Sameh Elnikety, Rodrigo Fonseca, and Adam Belay. Mak- ing kernel bypass practical for the cloud with junction. In21st USENIX Symposium on Networked Systems Design and Implementation (NSDI 24), pages 55–73, 2024

    work page 2024

  34. [34]

    Harmonizing performance and isolation in microker- nels with efficient intra-kernel isolation and communication

    Jinyu Gu, Xinyue Wu, Wentai Li, Nian Liu, Zeyu Mi, Yubin Xia, and Haibo Chen. Harmonizing performance and isolation in microker- nels with efficient intra-kernel isolation and communication. In2020 USENIX Annual Technical Conference (USENIX ATC 20), pages 401–417, 2020. 13 Wu et al., Yiyang Wu, Xunjie Wang, Jinyu Gu, and Haibo Chen

    work page 2020

  35. [35]

    A {Hardware-Software} co-design for efficient{Intra-Enclave} isolation

    Jinyu Gu, Bojun Zhu, Mingyu Li, Wentai Li, Yubin Xia, and Haibo Chen. A {Hardware-Software} co-design for efficient{Intra-Enclave} isolation. In31st USENIX Security Symposium (USENIX Security 22), pages 3129–3145, 2022

    work page 2022

  36. [36]

    Bulkhead: secure, scalable, and efficient kernel compart- mentalization with pks.arXiv preprint arXiv:2409.09606, 2024

    Yinggang Guo, Zicheng Wang, Weiheng Bai, Qingkai Zeng, and Kangjie Lu. Bulkhead: secure, scalable, and efficient kernel compart- mentalization with pks.arXiv preprint arXiv:2409.09606, 2024

    work page arXiv 2024

  37. [37]

    Sok: A com- prehensive analysis and evaluation of docker container attack and defense mechanisms

    Md Sadun Haq, Thien Duc Nguyen, Ali Şaman Tosun, Franziska Vollmer, Turgay Korkmaz, and Ahmad-Reza Sadeghi. Sok: A com- prehensive analysis and evaluation of docker container attack and defense mechanisms. In2024 IEEE symposium on security and privacy (SP), pages 4573–4590. IEEE, 2024

    work page 2024

  38. [38]

    Cross container attacks: The bewildered eBPF on clouds

    Yi He, Roland Guo, Yunlong Xing, Xijia Che, Kun Sun, Zhuotao Liu, Ke Xu, and Qi Li. Cross container attacks: The bewildered eBPF on clouds. In32nd USENIX Security Symposium (USENIX Security 23). USENIX Association, 2023

    work page 2023

  39. [39]

    Hodor:{Intra- Process} isolation for {High-Throughput} data plane libraries

    Mohammad Hedayati, Spyridoula Gravani, Ethan Johnson, John Criswell, Michael L Scott, Kai Shen, and Mike Marty. Hodor:{Intra- Process} isolation for {High-Throughput} data plane libraries. In 2019 USENIX Annual Technical Conference (USENIX ATC 19), pages 489–504, 2019

    work page 2019

  40. [40]

    Virtio-mem: Paravirtual- ized memory hot (un) plug

    David Hildenbrand and Martin Schulz. Virtio-mem: Paravirtual- ized memory hot (un) plug. InProceedings of the 17th ACM SIG- PLAN/SIGOPS International Conference on Virtual Execution Environ- ments, pages 1–14, 2021

    work page 2021

  41. [41]

    Everything you always wanted to know about plan- ning: (but were afraid to ask)

    Jörg Hoffmann. Everything you always wanted to know about plan- ning: (but were afraid to ask). InAnnual Conference on Artificial Intelligence, pages 1–13. Springer, 2011

    work page 2011

  42. [42]

    Pvm: Efficient shadow paging for deploying secure containers in cloud-native envi- ronment

    Hang Huang, Jiangshan Lai, Jia Rao, Hui Lu, Wenlong Hou, Hang Su, Quan Xu, Jiang Zhong, Jiahao Zeng, Xu Wang, et al. Pvm: Efficient shadow paging for deploying secure containers in cloud-native envi- ronment. InProceedings of the 29th Symposium on Operating Systems Principles, pages 515–530, 2023

    work page 2023

  43. [43]

    TrEnv: Transparently share serverless execution environments across different functions and nodes

    Jialiang Huang, Mingxing Zhang, Teng Ma, Zheng Liu, Sixing Lin, Kang Chen, Jinlei Jiang, Xia Liao, Yingdi Shan, Ning Zhang, Mengting Lu, Tao Ma, Haifeng Gong, and Yongwei Wu. TrEnv: Transparently share serverless execution environments across different functions and nodes. InProceedings of the ACM SIGOPS 30th Symposium on Operating Systems Principles (SOS...

    work page 2024

  44. [44]

    Memory protection keys for userspace, 2019

    Intel. Memory protection keys for userspace, 2019

    work page 2019

  45. [45]

    A container security survey: Exploits, attacks, and defenses.ACM Com- puting Surveys, 57(7):1–36, 2025

    Omar Jarkas, Ryan Ko, Naipeng Dong, and Redowan Mahmud. A container security survey: Exploits, attacks, and defenses.ACM Com- puting Surveys, 57(7):1–36, 2025

    work page 2025

  46. [46]

    Nightcore: efficient and scalable serverless computing for latency-sensitive, interactive microservices

    Zhipeng Jia and Emmett Witchel. Nightcore: efficient and scalable serverless computing for latency-sensitive, interactive microservices. InProceedings of the 26th ACM international conference on architectural support for programming languages and operating systems, pages 152– 166, 2021

    work page 2021

  47. [47]

    SWE-bench: Can Language Models Resolve Real-World GitHub Issues?

    Carlos E Jimenez, John Yang, Alexander Wettig, Shunyu Yao, Kexin Pei, Ofir Press, and Karthik Narasimhan. Swe-bench: Can language models resolve real-world github issues?arXiv preprint arXiv:2310.06770, 2023

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  48. [48]

    kvm: the linux virtual machine monitor

    Avi Kivity, Yaniv Kamay, Dor Laor, Uri Lublin, and Anthony Liguori. kvm: the linux virtual machine monitor. InProceedings of the Linux symposium, volume 1, pages 225–230. Dttawa, Dntorio, Canada, 2007

    work page 2007

  49. [49]

    Fhpm: Fine-grained huge page management for virtualization.arXiv preprint arXiv:2307.10618, 2023

    Chuandong Li, Sai Sha, Yangqing Zeng, Xiran Yang, Yingwei Luo, Xiaolin Wang, and Zhenlin Wang. Fhpm: Fine-grained huge page management for virtualization.arXiv preprint arXiv:2307.10618, 2023

    work page arXiv 2023

  50. [50]

    SkillsBench: Benchmarking How Well Agent Skills Work Across Diverse Tasks

    Xiangyi Li, Wenbo Chen, Yimin Liu, Shenghan Zheng, Xiaokun Chen, Yifeng He, Yubo Li, Bingran You, Haotian Shen, Jiankai Sun, et al. Skillsbench: Benchmarking how well agent skills work across diverse tasks.arXiv preprint arXiv:2602.12670, 2026

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  51. [51]

    {RunD}: A lightweight secure container runtime for high-density deployment and high-concurrency startup in serverless computing

    Zijun Li, Jiagan Cheng, Quan Chen, Eryu Guan, Zizheng Bian, Yi Tao, Bin Zha, Qiang Wang, Weidong Han, and Minyi Guo. {RunD}: A lightweight secure container runtime for high-density deployment and high-concurrency startup in serverless computing. In2022 USENIX Annual Technical Conference (USENIX ATC 22), pages 53–68, 2022

    work page 2022

  52. [52]

    Neve: Nested virtualization extensions for arm

    Jin Tack Lim, Christoffer Dall, Shih-Wei Li, Jason Nieh, and Marc Zyngier. Neve: Nested virtualization extensions for arm. InProceedings of the 26th Symposium on Operating Systems Principles, pages 201–217, 2017

    work page 2017

  53. [53]

    Optimizing nested virtualization perfor- mance using direct virtual hardware

    Jin Tack Lim and Jason Nieh. Optimizing nested virtualization perfor- mance using direct virtual hardware. InProceedings of the Twenty-Fifth International Conference on Architectural Support for Programming Languages and Operating Systems, pages 557–574, 2020

    work page 2020

  54. [54]

    Kit: Testing os-level virtualization for functional interference bugs

    Congyu Liu, Sishuai Gong, and Pedro Fonseca. Kit: Testing os-level virtualization for functional interference bugs. InProceedings of the 28th ACM International Conference on Architectural Support for Pro- gramming Languages and Operating Systems, Volume 2, pages 427–441, 2023

    work page 2023

  55. [55]

    Nanozone: Scalable, efficient, and secure memory protection for arm cca.arXiv preprint arXiv:2506.07034, 2025

    Shiqi Liu, Yongpeng Gao, Mingyang Zhang, and Jie Wang. Nanozone: Scalable, efficient, and secure memory protection for arm cca.arXiv preprint arXiv:2506.07034, 2025

    work page arXiv 2025

  56. [56]

    Transcendent memory and linux

    Dan Magenheimer, Chris Mason, Dave McCracken, and Kurt Hackel. Transcendent memory and linux. InProceedings of the Linux Sympo- sium, pages 191–200, 2009

    work page 2009

  57. [57]

    {LITESHIELD}: Secure containers via lightweight, composable userspace {𝜇 Kernel} services

    Kaesi Manakkal, Nathan Daughety, Marcus Pendleton, and Hui Lu. {LITESHIELD}: Secure containers via lightweight, composable userspace {𝜇 Kernel} services. In2025 USENIX Annual Technical Con- ference (USENIX ATC 25), pages 973–985, 2025

    work page 2025

  58. [58]

    My vm is lighter (and safer) than your container

    Filipe Manco, Costin Lupu, Florian Schmidt, Jose Mendes, Simon Kuen- zer, Sumit Sati, Kenichi Yasukata, Costin Raiciu, and Felipe Huici. My vm is lighter (and safer) than your container. InProceedings of the 26th Symposium on Operating Systems Principles, pages 218–233, 2017

    work page 2017

  59. [59]

    Arpaci-Dusseau, and Remzi H

    Edward Oakes, Leon Yang, Dennis Zhou, Kevin Houck, Tyler Harter, Andrea C. Arpaci-Dusseau, and Remzi H. Arpaci-Dusseau. SOCK: Rapid task provisioning with serverless-optimized containers. In2018 USENIX Annual Technical Conference (USENIX ATC 18), pages 57–70. USENIX Association, 2018

    work page 2018

  60. [60]

    Fast local page-tables for virtualized numa servers with vmitosis

    Ashish Panwar, Reto Achermann, Arkaprava Basu, Abhishek Bhat- tacharjee, K Gopinath, and Jayneel Gandhi. Fast local page-tables for virtualized numa servers with vmitosis. InProceedings of the 26th ACM International Conference on Architectural Support for Programming Languages and Operating Systems, pages 194–210, 2021

    work page 2021

  61. [61]

    libmpk: Software abstraction for intel memory protection keys (intel {MPK})

    Soyeon Park, Sangho Lee, Wen Xu, Hyungon Moon, and Taesoo Kim. libmpk: Software abstraction for intel memory protection keys (intel {MPK}). In2019 USENIX Annual Technical Conference (USENIX ATC 19), pages 241–254, 2019

    work page 2019

  62. [62]

    Spright: extracting the server from serverless computing! high-performance ebpf-based event-driven, shared-memory process- ing

    Shixiong Qi, Leslie Monis, Ziteng Zeng, Ian-chin Wang, and KK Ra- makrishnan. Spright: extracting the server from serverless computing! high-performance ebpf-based event-driven, shared-memory process- ing. InProceedings of the ACM SIGCOMM 2022 Conference, pages 780–794, 2022

    work page 2022

  63. [63]

    Resizing memory with balloons and hotplug

    Joel H Schopp, Keir Fraser, and Martine J Silbermann. Resizing memory with balloons and hotplug. InProceedings of the Linux Symposium, volume 2, pages 313–319, 2006

    work page 2006

  64. [64]

    Nanvix: A Multikernel OS Design for High-Density Serverless Deployments

    Carlos Segarra, Pedro Henrique Penna, Enrique Saurez, Íñigo Goiri, Peter Pietzuch, Shan Lu, and Rodrigo Fonseca. Nanvix: A multikernel os design for high-density serverless deployments.arXiv preprint arXiv:2604.11669, 2026

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  65. [65]

    X-containers: Breaking down barriers to improve performance and iso- lation of cloud-native containers

    Zhiming Shen, Zhen Sun, Gur-Eyal Sela, Eugene Bagdasaryan, Christina Delimitrou, Robbert Van Renesse, and Hakim Weatherspoon. X-containers: Breaking down barriers to improve performance and iso- lation of cloud-native containers. InProceedings of the Twenty-Fourth International Conference on Architectural Support for Programming Languages and Operating Sy...

    work page 2019

  66. [66]

    A hardware-software co-design for efficient secure containers

    Jiacheng Shi, Yang Yu, Jinyu Gu, and Yubin Xia. A hardware-software co-design for efficient secure containers. InProceedings of the Twentieth European Conference on Computer Systems, pages 1229–1245, 2025. 14 ParaCell: Paravirtualized Secure Containers with Lightweight Intra-Container Isolation and Intent-Driven Memory Management

    work page 2025

  67. [67]

    Locked in, leaked out: Measuring isolation via kernel locks.arXiv preprint arXiv:2507.21248, 2025

    Michael M Swift et al. Locked in, leaked out: Measuring isolation via kernel locks.arXiv preprint arXiv:2507.21248, 2025

    work page arXiv 2025

  68. [68]

    A study on container vulnerability exploit detection

    Olufogorehan Tunde-Onadele, Jingzhu He, Ting Dai, and Xiaohui Gu. A study on container vulnerability exploit detection. In2019 ieee international conference on cloud engineering (IC2E), pages 121–127. IEEE, 2019

    work page 2019

  69. [69]

    Intel virtualization technology.Computer, 38(5):48–56, 2005

    Rich Uhlig, Gil Neiger, Dion Rodgers, Amy L Santoni, Fernando CM Martins, Andrew V Anderson, Steven M Bennett, Alain Kagi, Felix H Leung, and Larry Smith. Intel virtualization technology.Computer, 38(5):48–56, 2005

    work page 2005

  70. [70]

    Duarte, Michael Sammler, Peter Druschel, and Deepak Garg

    Anjo Vahldiek-Oberwagner, Eslam Elnikety, Nuno O. Duarte, Michael Sammler, Peter Druschel, and Deepak Garg. Erim: secure, efficient in- process isolation with protection keys (mpk). InProceedings of the 28th USENIX Conference on Security Symposium, SEC’19, page 1221–1238, USA, 2019. USENIX Association

    work page 2019

  71. [71]

    Using smt to accel- erate nested virtualization

    Lluís Vilanova, Nadav Amit, and Yoav Etsion. Using smt to accel- erate nested virtualization. InProceedings of the 46th International Symposium on Computer Architecture, pages 750–761, 2019

    work page 2019

  72. [72]

    Secure virtual machine architecture reference man- ual.AMD Publication, 33047, 2005

    A Virtualization. Secure virtual machine architecture reference man- ual.AMD Publication, 33047, 2005

    work page 2005

  73. [73]

    Detecting noisy neighbors in cpu-isolated cgroups environments

    Simon Volpert, Sascha Winkelhofer, Jörg Domaschka, and Stefan Wes- ner. Detecting noisy neighbors in cpu-isolated cgroups environments. InProceedings of the 16th ACM/SPEC International Conference on Per- formance Engineering, ICPE ’25, page 224–231, New York, NY, USA,

    work page

  74. [74]

    Association for Computing Machinery

    work page

  75. [75]

    You shall not (by) pass! practical, secure, and fast pku-based sandboxing

    Alexios Voulimeneas, Jonas Vinck, Ruben Mechelinck, and Stijn Vol- ckaert. You shall not (by) pass! practical, secure, and fast pku-based sandboxing. InProceedings of the Seventeenth European Conference on Computer Systems, pages 266–282, 2022

    work page 2022

  76. [76]

    Memory resource management in vmware esx server.ACM SIGOPS Operating Systems Review, 36(SI):181–194, 2002

    Carl A Waldspurger. Memory resource management in vmware esx server.ACM SIGOPS Operating Systems Review, 36(SI):181–194, 2002

    work page 2002

  77. [77]

    Characterizing and optimizing kernel resource isolation for containers.Future Generation Computer Systems, 141:218–229, 2023

    Kun Wang, Song Wu, Kun Suo, Yijie Liu, Hang Huang, Zhuo Huang, and Hai Jin. Characterizing and optimizing kernel resource isolation for containers.Future Generation Computer Systems, 141:218–229, 2023

    work page 2023

  78. [78]

    Efficient memory overcom- mitment for {I/O} passthrough enabled {VMs} via fine-grained page meta-data management

    Yaohui Wang, Ben Luo, and Yibin Shen. Efficient memory overcom- mitment for {I/O} passthrough enabled {VMs} via fine-grained page meta-data management. In2023 USENIX Annual Technical Conference (USENIX ATC 23), pages 769–783, 2023

    work page 2023

  79. [79]

    No provisioned concurrency: Fast RDMA- codesigned remote fork for serverless computing

    Xingda Wei, Fangming Lu, Tianxia Wang, Jinyu Gu, Yuhan Yang, Rong Chen, and Haibo Chen. No provisioned concurrency: Fast RDMA- codesigned remote fork for serverless computing. In17th USENIX Symposium on Operating Systems Design and Implementation (OSDI 23), pages 497–517. USENIX Association, 2023

    work page 2023

  80. [80]

    Enabling efficient hypervisor- as-a-service clouds with ephemeral virtualization

    Dan Williams, Yaohui Hu, Umesh Deshpande, Piush K Sinha, Nilton Bila, Kartik Gopalan, and Hani Jamjoom. Enabling efficient hypervisor- as-a-service clouds with ephemeral virtualization. InProceedings of the12th ACM SIGPLAN/SIGOPS International Conference on Virtual Execution Environments, pages 79–92, 2016

    work page 2016

Showing first 80 references.