pith. machine review for the scientific record. sign in

arxiv: 2604.11669 · v1 · submitted 2026-04-13 · 💻 cs.OS · cs.DC

Recognition: unknown

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

Carlos Segarra, Enrique Saurez, \'I\~nigo Goiri, Pedro Henrique Penna, Peter Pietzuch, Rodrigo Fonseca, Shan Lu

Authors on Pith no claims yet

Pith reviewed 2026-05-10 16:32 UTC · model grok-4.3

classification 💻 cs.OS cs.DC
keywords serverless computingmultikernel OSdeployment densityvirtual machine isolationI/O multiplexingmicrokernel designtenant isolation
0
0 comments X

The pith

Nanvix separates ephemeral per-invocation state from shared tenant state to run far more serverless applications per server.

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

The paper presents Nanvix as a multikernel operating system for serverless environments that keeps state unique to each application invocation separate from state shared across multiple invocations of the same tenant. Each application runs inside its own lightweight user virtual machine equipped with a micro-kernel for threads and memory management, while all input and output operations are forwarded to a single system virtual machine per tenant that runs a full macro-kernel with device drivers. This arrangement delivers hypervisor-level isolation between different tenants and reduces resource contention within a tenant by centralizing I/O handling. The design produces substantially faster application startup and allows the same workload to run on many fewer physical servers. Readers would care because higher deployment density directly lowers the hardware required to support serverless workloads while preserving security boundaries.

Core claim

Nanvix disaggregates ephemeral execution state, unique per application invocation, from long-lived persistent state shared among invocations from the same tenant. Applications execute inside a lightweight user VM running a micro-kernel that implements threads and memory and forwards all I/O requests to a system VM. The system VM runs a macro-kernel with device drivers and is shared among all invocations from the same tenant. The split design achieves strong hypervisor isolation across tenants without sacrificing application performance and reduces same-tenant contention by multiplexing all I/O requests to the system VM, yielding order-of-magnitude lower startup times and the ability to serve

What carries the argument

The per-invocation user VM with micro-kernel paired with a per-tenant system VM with macro-kernel that receives all forwarded I/O.

If this is right

  • Application startup times drop by roughly an order of magnitude with only moderate I/O overhead.
  • A production trace replay requires 20-100 times fewer host servers than current systems.
  • Strong isolation between tenants is preserved through the hypervisor layer.
  • Same-tenant contention is lowered by centralizing I/O in the shared system VM.
  • Overall serverless deployment density increases without added hardware.

Where Pith is reading between the lines

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

  • Similar state disaggregation could be applied to other cloud execution models where isolation and density trade off.
  • If the I/O multiplexing scales, it opens a path to denser multi-tenant runtimes beyond serverless functions.
  • An experiment running identical workloads on both Nanvix and conventional container or VM stacks would quantify the density gain directly.
  • The design may reduce the total cost of ownership for serverless platforms by allowing more work per physical machine.

Load-bearing premise

That routing all I/O from many concurrent invocations of the same tenant through one shared system VM will avoid both performance collapse and new side-channel vulnerabilities.

What would settle it

Measure I/O latency and side-channel leakage when running a high-concurrency trace from a single tenant on Nanvix versus an equivalent number of fully isolated per-invocation VMs; a large degradation or new leaks would falsify the density claim.

Figures

Figures reproduced from arXiv: 2604.11669 by Carlos Segarra, Enrique Saurez, \'I\~nigo Goiri, Pedro Henrique Penna, Peter Pietzuch, Rodrigo Fonseca, Shan Lu.

Figure 1
Figure 1. Figure 1: Serverless design space (Serverless providers wish to optimize deployment density as a proxy for resource efficiency, but struggle to do so while maintaining low cold start latencies, strong inter-tenant isolation, and application compatibility.) service level objective (SLO) targets [33]. Unfortunately, it is hard to achieve deployment density while maintaining low cold start latencies and strong isolatio… view at source ↗
Figure 2
Figure 2. Figure 2: Nanvix is a multikernel OS for serverless (Ex￾isting multi-tenant serverless platforms offer compatibility by including all OS functionality in the guest, fundamentally limiting performance and density. Nanvix, on the other hand, advocates for a disaggregated design.) but sacrifice application compatibility by imposing substan￾tially different programming and execution models [28, 54]. Unfortunately, the s… view at source ↗
Figure 4
Figure 4. Figure 4: Containers in VM deployment (This approach employs a base VM that can either have a static allocation of resources, in which case some resources may be stranded, or may grow and shrink dynamically, adding latency to the request’s critical path.) vCPU 1 4 8 – – – 1 4 8 Mem (MiB) – – – 128 512 1024 128 512 1024 p50 (ms) 73.5 94 141 22 22.5 25 77 86 141 p99 (ms) 104 226 490 24 24.9 36.1 89.5 108 380 [PITH_FU… view at source ↗
Figure 5
Figure 5. Figure 5: Overview of Nanvix (Nanvix adopts a multiker￾nel design where a micro-kernel executes application code and a macro-kernel multiplexes I/O requests from user VMs belonging to the same tenant.) Low cold start latency. With high probability (see [PITH_FULL_IMAGE:figures/full_fig_p004_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: provides a high-level overview of the user VM archi￾tecture, contrasting it with state-of-the-art VMMs capable of hosting Linux [1, 20], and illustrates the process of executing a read POSIX call. Application code in Nanvix is executed by one or many User threads in the micro-kernel ( 1 , [PITH_FULL_IMAGE:figures/full_fig_p005_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: System VM architecture (The system VM exe￾cutes inter-kernel calls from user VMs belonging to the same tenant and interfaces with the different client applications.) contain nor execute any tenant-provided application code, and its image is generic and shared across all tenants. This means that system VMs, in spite of running a full-weight kernel, can be easily pre-warmed and pooled, thus eliminat￾ing the … view at source ↗
Figure 8
Figure 8. Figure 8: Nanvix deployment (Groups of user VMs and a system VM belonging to the same tenant are deployed inside a network namespace, and user VMs that do not require any I/O can be deployed in standalone mode.) User VM. The micro-kernel and the VMM are implemented in 17 and 11 kLOC, respectively. The VMM has support for the KVM hypervisor and preliminary support for the Win￾dows Hypervisor Platform (WHP). Nanvix ca… view at source ↗
Figure 9
Figure 9. Figure 9: Deployment density (We spawn sandboxes in a closed-loop until we consume 1 GiB of system memory, and measure each sandbox’s contribution by reading MemAvailable from /proc/meminfo. Each dot represents the median cold start latency and memory footprint across runs. The sandbox count is the minimum number of sandboxes, across all 10 runs, we can fit in 1 GiB.) Baseline p50 (us) p99 (us) Slowdown Firecracker … view at source ↗
Figure 10
Figure 10. Figure 10: Throughput characteristic (We present the p50, marker, and p99, faded. See Tab. 5 for the legend.) we report p50 and p99 across a million iterations and the slowdown compared to Nanvix. The results in Tab. 5 show that Nanvix introduces, at worst, a 50% overhead on round-trip latency compared to the other virtualized baselines. Our profiling indicates that this over￾head is not due to the additional hop, b… view at source ↗
Figure 11
Figure 11. Figure 11: Trace replay (We replay the Huawei trace and measure the per-request p50, marker, and p99, shade, latency. See the table at the right-hand side for the legend.) Baseline Peak RPS Mem (MiB) # Servers Firecracker 43.1 2211 100 20x Cloud Hypervisor 43.1 5479 100 20x Unikraft 43.1 1617 100 20x gVisor 8.62 151 500 100x Firecracker-S 43.1 382 100 20x CloudHypervisor-S 43.1 125 100 20x Hyperlight 862 230 5 1x Pr… view at source ↗
read the original abstract

Serverless providers strive for high resource utilization by optimizing deployment density: how many applications can be deployed per host server. However, achieving high deployment density without compromising application performance or isolation remains an open challenge. High density can be achieved by sharing components across applications, yet applications from different tenants must be strongly isolated from each other due to the risk of side-channel attacks. Sharing components across applications from the same tenant, if done naively, can introduce contention on host resources thus negatively affecting application performance. We describe Nanvix, a new multikernel OS that disaggregates ephemeral execution state, unique per application invocation, from long-lived persistent state, shared among invocations from the same tenant. Applications in Nanvix execute inside a lightweight user VM running a micro-kernel that implements threads and memory, and forwards all I/O requests to a system VM. The system VM runs a macro-kernel with a rich set of device drivers and is shared among all invocations from the same tenant. Nanvix' split design achieves strong hypervisor isolation across tenants without sacrificing application performance, and reduces same-tenant contention by multiplexing all I/O requests to the system VM. Thanks to a system-wide co-design, Nanvix achieves order-of-magnitude lower application start up times with moderate I/O overheads. When replaying a production trace, Nanvix needs 20-100x fewer host servers compared to state-of-the-art systems, improving deployment density

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

3 major / 2 minor

Summary. The paper describes Nanvix, a multikernel OS for high-density serverless deployments. It separates ephemeral execution state per application invocation into lightweight user VMs running micro-kernels from persistent state shared among same-tenant invocations in a system VM running a macro-kernel. All I/O is forwarded to the system VM to maintain isolation across tenants while aiming to reduce intra-tenant contention. The design is claimed to achieve order-of-magnitude lower startup times with moderate I/O overheads and 20-100x higher deployment density on production trace replays compared to state-of-the-art systems.

Significance. Should the claims be supported by detailed and reproducible evaluations, this work could have substantial impact on serverless platforms by significantly increasing deployment density while upholding strong isolation. The multikernel disaggregation tailored to serverless invocation patterns is a novel contribution, and the use of production traces for validation is commendable.

major comments (3)
  1. Abstract: Performance claims such as 'order-of-magnitude lower application start up times' and '20-100x fewer host servers' are made without reference to specific evaluation methodologies, baselines, or data, which are critical for assessing the central claims of improved density.
  2. §3 (Architecture): The architecture forwards every I/O request to a single shared system VM per tenant; however, there is no quantitative evaluation of potential contention or latency under high concurrency scenarios from the same tenant, which underpins the assertion that this reduces contention and enables the reported performance gains.
  3. §5 (Evaluation): The production trace replay experiment does not detail the state-of-the-art systems compared against, the specific metrics for density (e.g., how server count is determined), or any error analysis, making it hard to verify the 20-100x improvement.
minor comments (2)
  1. Abstract: Consider expanding the abstract to include a sentence on the specific hypervisor or kernel implementations used for better context.
  2. Throughout the manuscript: Clarify the distinction between 'lightweight user VM' and 'system VM' with consistent definitions early on.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive and detailed review. We address each major comment below and will revise the manuscript accordingly to improve clarity and completeness.

read point-by-point responses

  1. Referee: Abstract: Performance claims such as 'order-of-magnitude lower application start up times' and '20-100x fewer host servers' are made without reference to specific evaluation methodologies, baselines, or data, which are critical for assessing the central claims of improved density.

    Authors: We agree that the abstract would benefit from brief references to the supporting evaluation. In the revised manuscript we will update the abstract to note that startup-time claims are based on microbenchmarks and that the density improvement is measured via replay of a production trace against state-of-the-art serverless runtimes. Full methodologies, baselines, and quantitative data remain in Sections 4 and 5. revision: yes

  2. Referee: §3 (Architecture): The architecture forwards every I/O request to a single shared system VM per tenant; however, there is no quantitative evaluation of potential contention or latency under high concurrency scenarios from the same tenant, which underpins the assertion that this reduces contention and enables the reported performance gains.

    Authors: The design forwards I/O to the tenant-shared system VM to enable centralized multiplexing and lower per-invocation overhead. We acknowledge the absence of quantitative contention measurements under high same-tenant concurrency. We will add targeted experiments in the evaluation section that vary the number of concurrent invocations per tenant and report I/O latency and throughput to demonstrate the contention-reduction benefit. revision: yes

  3. Referee: §5 (Evaluation): The production trace replay experiment does not detail the state-of-the-art systems compared against, the specific metrics for density (e.g., how server count is determined), or any error analysis, making it hard to verify the 20-100x improvement.

    Authors: We will expand Section 5 to explicitly name the compared systems and their configurations, define the density metric as the minimum number of host servers needed to replay the trace while keeping tail latencies within SLA bounds, and include error analysis with standard deviations from repeated runs. These additions will make the 20-100x claim directly verifiable. revision: yes

Circularity Check

0 steps flagged

No circularity: claims rest on architecture description and empirical trace replay

full rationale

The paper describes a multikernel OS design that disaggregates per-invocation state from per-tenant persistent state, with I/O forwarded to a shared system VM. Performance claims (order-of-magnitude lower startup times, 20-100x fewer servers on production trace replay) are presented as outcomes of this co-design and evaluated via replay rather than derived from equations or parameters. No self-definitional loops, fitted inputs renamed as predictions, or load-bearing self-citations appear in the provided text. The central results are externally falsifiable through the described evaluation methodology and do not reduce to quantities defined by the authors' own prior results.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 2 invented entities

The design rests on standard OS isolation assumptions plus two paper-specific premises about contention and multiplexing; no free parameters or externally validated invented entities are described.

axioms (2)
  • domain assumption Strong hypervisor isolation is required between tenants because of side-channel attack risks.
    Explicitly stated as motivation in the abstract.
  • ad hoc to paper Multiplexing I/O through a shared per-tenant system VM avoids contention that would otherwise degrade performance.
    Central design choice whose validity determines the density claim.
invented entities (2)
  • Lightweight user VM running a micro-kernel no independent evidence
    purpose: Execute threads and memory management for a single application invocation
    New per-invocation component introduced to keep ephemeral state private and fast.
  • Shared system VM running a macro-kernel no independent evidence
    purpose: Provide device drivers and I/O services to all invocations from the same tenant
    New shared component introduced to reduce same-tenant contention.

pith-pipeline@v0.9.0 · 5588 in / 1395 out tokens · 40064 ms · 2026-05-10T16:32:36.630534+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

59 extracted references · 1 canonical work pages

  1. [1]

    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 12 USENIX Symposium on Networked Systems Design and Implementation, NSDI, 2020

    2020

  2. [2]

    Drops: Managing serverless resource pools in microsoft azure functions

    Ahmed Alquraan, Abdelrahman Baba, Rafael Mendes, Sameh Elnikety, Paul Batum, Yan Chen, Seth Safi, Hamid Henry Fine, and Samer Al- Kiswany. Drops: Managing serverless resource pools in microsoft azure functions. InProceedings of the Twentyfirst European Conference on Computer Systems, EuroSys ’26, 2026

    2026

  3. [3]

    Improving startup performance with Lambda SnapStart.https: //docs.aws.amazon.com/lambda/latest/dg/snapstart.html, 2026

    AWS. Improving startup performance with Lambda SnapStart.https: //docs.aws.amazon.com/lambda/latest/dg/snapstart.html, 2026

    2026

  4. [4]

    Lambda - Provisioned Concurrency.https://docs.aws.amazon

    AWS. Lambda - Provisioned Concurrency.https://docs.aws.amazon. com/lambda/latest/dg/provisioned-concurrency.html, 2026

    2026

  5. [5]

    Lambda - Runtime Environment.https://docs.aws.amazon.com/ lambda/latest/dg/lambda-runtime-environment.html, 2026

    AWS. Lambda - Runtime Environment.https://docs.aws.amazon.com/ lambda/latest/dg/lambda-runtime-environment.html, 2026

    2026

  6. [6]

    Wei Bai, Shanim Sainul Abdeen, Ankit Agrawal, Krishan Kumar Attre, Paramvir Bahl, Ameya Bhagat, Gowri Bhaskara, Tanya Brokhman, Lei Cao, Ahmad Cheema, Rebecca Chow, Jeff Cohen, Mahmoud Elhaddad, Vivek Ette, Igal Figlin, Daniel Firestone, Mathew George, Ilya German, Lakhmeet Ghai, Eric Green, Albert Greenberg, Manish Gupta, Randy Haagens, Matthew Hendel, R...

    2023

  7. [7]

    The multikernel: a new os architecture for scalable multicore systems

    Andrew Baumann, Paul Barham, Pierre-Evariste Dagand, Tim Harris, Rebecca Isaacs, Simon Peter, Timothy Roscoe, Adrian Schüpbach, and Akhilesh Singhania. The multikernel: a new os architecture for scalable multicore systems. InProceedings of the ACM SIGOPS 22nd Symposium on Operating Systems Principles, SOSP ’09, 2009

    2009

  8. [8]

    Dedup est machina: Memory deduplication as an advanced exploita- tion vector

    Erik Bosman, Kaveh Razavi, Herbert Bos, and Cristiano Giuffrida. Dedup est machina: Memory deduplication as an advanced exploita- tion vector. In2016 IEEE Symposium on Security and Privacy, SP ’16, 2016

    2016

  9. [9]

    SEUSS: skip redundant paths to make server- less fast

    James Cadden, Thomas Unger, Yara Awad, Han Dong, Orran Krieger, and Jonathan Appavoo. SEUSS: skip redundant paths to make server- less fast. InEuroSys, 2020

    2020

  10. [10]

    Fork in the road: reflections and optimizations for cold start latency in production serverless systems

    Xiaohu Chai, Tianyu Zhou, Keyang Hu, Jianfeng Tan, Tiwei Bie, Anqi Shen, Dawei Shen, Qi Xing, Shun Song, Tongkai Yang, Le Gao, Feng Yu, Zhengyu He, Dong Du, Yubin Xia, Kang Chen, and Yu Chen. Fork in the road: reflections and optimizations for cold start latency in production serverless systems. InProceedings of the 19th USENIX Conference on Operating Sys...

    2025

  11. [11]

    Run.https://cloud.google.com/run, 2026

    Google Cloud. Run.https://cloud.google.com/run, 2026

    2026

  12. [12]

    The cloud that works for developers, not the other way around.https://workers.cloudflare.com/, 2026

    CloudFlare. The cloud that works for developers, not the other way around.https://workers.cloudflare.com/, 2026

    2026

  13. [13]

    ACPI.https://wiki.osdev.org/ACPI, 2026

    OS Dev. ACPI.https://wiki.osdev.org/ACPI, 2026

    2026

  14. [14]

    VirtIO.https://wiki.osdev.org/Virtio, 2026

    OS Dev. VirtIO.https://wiki.osdev.org/Virtio, 2026

    2026

  15. [15]

    Handling uniqueness with Lambda SnapStart.https://docs

    AWS Docs. Handling uniqueness with Lambda SnapStart.https://docs. aws.amazon.com/lambda/latest/dg/snapstart-uniqueness.html, 2026

    2026

  16. [16]

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

    Dong Du, Tianyi Yu, Yubin Xia, Binyu Zang, Guanglu Yan, Chenggang Qin, Qixuan Wu, and Haibo Chen. Catalyzer: Sub-millisecond startup for serverless computing with initialization-less booting. InProceedings of the Twenty-Fifth International Conference on Architectural Support for Programming Languages and Operating Systems, ASPLOS ’20, 2020

    2020

  17. [17]

    Open source Kubernetes-native Serverless Framework.https: //fission.io, 2026

    Fission. Open source Kubernetes-native Serverless Framework.https: //fission.io, 2026

    2026

  18. [18]

    Making 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. Making kernel bypass practical for the cloud with Junction. InNSDI, 2024

    2024

  19. [19]

    Mem- ory matters: Load-time deduplication for unikernels

    Gaulthier Gain, Benoît Knott, Cyril Soldani, and Laurent Mathy. Mem- ory matters: Load-time deduplication for unikernels. InProceedings of the 2025 ACM Symposium on Cloud Computing, SoCC ’25, 2026

    2025

  20. [20]

    Run Cloud Virtual Machines Securely and Effi- ciently.https://www.cloudhypervisor.org/, 2026

    Cloud Hypervisor. Run Cloud Virtual Machines Securely and Effi- ciently.https://www.cloudhypervisor.org/, 2026

    2026

  21. [21]

    Serverless cold starts and where to find them

    Artjom Joosen, Ahmed Hassan, Martin Asenov, Rajkarn Singh, Luke Darlow, Jianfeng Wang, Qiwen Deng, and Adam Barker. Serverless cold starts and where to find them. InProceedings of the Twentieth European Conference on Computer Systems, EuroSys ’25, 2025

    2025

  22. [22]

    Linefs: Efficient smartnic offload of a distributed file system with pipeline parallelism

    Jongyul Kim, Insu Jang, Waleed Reda, Jaeseong Im, Marco Canini, De- jan Kostić, Youngjin Kwon, Simon Peter, and Emmett Witchel. Linefs: Efficient smartnic offload of a distributed file system with pipeline parallelism. In28th Symposium on Operating Systems Principles, SOSP ’21, 2021

    2021

  23. [23]

    sel4: formal verification of an os kernel

    Gerwin Klein, Kevin Elphinstone, Gernot Heiser, June Andronick, David Cock, Philip Derrin, Dhammika Elkaduwe, Kai Engelhardt, Rafal Kolanski, Michael Norrish, Thomas Sewell, Harvey Tuch, and Simon Winwood. sel4: formal verification of an os kernel. InProceedings of the ACM SIGOPS 22nd Symposium on Operating Systems Principles, SOSP ’09, 2009

    2009

  24. [24]

    Spectre attacks: Exploit- ing speculative execution

    Paul Kocher, Jann Horn, Anders Fogh, , Daniel Genkin, Daniel Gruss, Werner Haas, Mike Hamburg, Moritz Lipp, Stefan Mangard, Thomas Prescher, Michael Schwarz, and Yuval Yarom. Spectre attacks: Exploit- ing speculative execution. In40th IEEE Symposium on Security and Privacy, S&P’19, 2019

    2019

  25. [25]

    Pronghorn: Effective checkpoint orchestration for serverless hot-starts

    Sumer Kohli, Shreyas Kharbanda, Rodrigo Bruno, Joao Carreira, and Pedro Fonseca. Pronghorn: Effective checkpoint orchestration for serverless hot-starts. InProceedings of the Nineteenth European Con- ference on Computer Systems, EuroSys ’24, 2024

    2024

  26. [26]

    Melding the serverless control plane with the conventional cluster manager for speed and resource efficiency.https: //arxiv.org/abs/2505.24551, 2026

    Leonid Kondrashov, Lazar Cvetković, Hancheng Wang, Boxi Zhou, and Dmitrii Ustiugov. Melding the serverless control plane with the conventional cluster manager for speed and resource efficiency.https: //arxiv.org/abs/2505.24551, 2026

    work page arXiv 2026

  27. [27]

    ufork: Sup- porting posix fork within a single-address-space os

    John Alistair Kressel, Hugo Lefeuvre, and Pierre Olivier. ufork: Sup- porting posix fork within a single-address-space os. InProceedings of the ACM SIGOPS 31st Symposium on Operating Systems Principles, SOSP ’25, 2025

    2025

  28. [28]

    Unlocking true elasticity for the cloud-native era with dandelion

    Tom Kuchler, Pinghe Li, Yazhuo Zhang, Lazar Cvetković, Boris Gora- nov, Tobias Stocker, Leon Thomm, Simone Kalbermatter, Tim Notter, Andrea Lattuada, and Ana Klimovic. Unlocking true elasticity for the cloud-native era with dandelion. InProceedings of the ACM SIGOPS 31st Symposium on Operating Systems Principles, SOSP ’25, 2025

    2025

  29. [29]

    Unikraft: Fast, specialized unikernels the easy way

    Simon Kuenzer, Vlad-Andrei Bădoiu, Hugo Lefeuvre, Sharan San- thanam, Alexander Jung, Gaulthier Gain, Cyril Soldani, Costin Lupu, Ştefan Teodorescu, Costi Răducanu, et al. Unikraft: Fast, specialized unikernels the easy way. InEuroSys, 2021

    2021

  30. [30]

    Squeezy: Rapid vm memory reclama- tion for serverless functions

    Orestis Lagkas Nikolos, Clhoe Alverti, Stratos Psomadakis, Georgios Goumas, and Nectarios Koziris. Squeezy: Rapid vm memory reclama- tion for serverless functions. InProceedings of the Twentyfirst European Conference on Computer Systems, EuroSys ’26, 2026

    2026

  31. [31]

    Lorch, Oded Padon, and Bryan Parno

    Andrea Lattuada, Travis Hance, Jay Bosamiya, Matthias Brun, Chanhee Cho, Hayley LeBlanc, Pranav Srinivasan, Reto Achermann, Tej Chajed, Chris Hawblitzel, Jon Howell, Jacob R. Lorch, Oded Padon, and Bryan Parno. Verus: A practical foundation for systems verification. In Proceedings of the ACM SIGOPS 30th Symposium on Operating Systems Principles, SOSP ’24, 2024

    2024

  32. [32]

    Loupe: Driving the development of os compatibility layers

    Hugo Lefeuvre, Gaulthier Gain, Vlad-Andrei Bădoiu, Daniel Dinca, Vlad-Radu Schiller, Costin Raiciu, Felipe Huici, and Pierre Olivier. Loupe: Driving the development of os compatibility layers. InProceed- ings of the 29th ACM International Conference on Architectural Support for Programming Languages and Operating Systems, Volume 1, ASPLOS 13 ’24, 2024

    2024

  33. [33]

    Lightweight and holistic-scalable serverless secure container runtime for high- density deployment and high-concurrency startup.IEEE Transactions on Computers, 2025

    Zijun Li, Chenyang Wu, Chuhao Xu, Quan Chen, Shuo Quan, Bin Zha, Qiang Wang, Weidong Han, Jie Wu, and Minyi Guo. Lightweight and holistic-scalable serverless secure container runtime for high- density deployment and high-concurrency startup.IEEE Transactions on Computers, 2025

    2025

  34. [34]

    Improving ipc by kernel design

    Jochen Liedtke. Improving ipc by kernel design. InProceedings of the Fourteenth ACM Symposium on Operating Systems Principles, SOSP ’93, 1993

    1993

  35. [35]

    Meltdown: Reading kernel memory from user space

    Moritz Lipp, Michael Schwarz, Daniel Gruss, Thomas Prescher, Werner Haas, Anders Fogh, Jann Horn, Stefan Mangard, Paul Kocher, Daniel Genkin, Yuval Yarom, and Mike Hamburg. Meltdown: Reading kernel memory from user space. In27th USENIX Security Symposium, USENIX Security 18, 2018

    2018

  36. [36]

    Unikernels: library operating systems for the cloud

    Anil Madhavapeddy, Richard Mortier, Charalampos Rotsos, David Scott, Balraj Singh, Thomas Gazagnaire, Steven Smith, Steven Hand, and Jon Crowcroft. Unikernels: library operating systems for the cloud. InProceedings of the Eighteenth International Conference on Ar- chitectural Support for Programming Languages and Operating Systems, ASPLOS ’13, 2013

    2013

  37. [37]

    Entropy for Clones.https: //github.com/firecracker-microvm/firecracker/blob/main/docs/ snapshotting/random-for-clones.md, 2026

    Firecracker Maintainers. Entropy for Clones.https: //github.com/firecracker-microvm/firecracker/blob/main/docs/ snapshotting/random-for-clones.md, 2026

    2026

  38. [38]

    Introducing Hyperlight: Virtual machine-based security for functions at scale.https://opensource.microsoft.com/blog/2024/11/ 07/introducing-hyperlight/, 2024

    Microsoft. Introducing Hyperlight: Virtual machine-based security for functions at scale.https://opensource.microsoft.com/blog/2024/11/ 07/introducing-hyperlight/, 2024

    2024

  39. [39]

    Metteagle: costs and benefits of implementing containers on microkernels

    Till Miemietz, Viktor Reusch, Matthias Hille, Lars Wrenger, Jana Eisoldt, Jan Klötzke, Max Kurze, Adam Lackorzynski, Michael Roitzsch, and Hermann Härtig. Metteagle: costs and benefits of implementing containers on microkernels. InProceedings of the 19th USENIX Con- ference on Operating Systems Design and Implementation, OSDI ’25, 2025

    2025

  40. [40]

    Nightingale, Orion Hodson, Ross McIlroy, Chris Hawblitzel, and Galen Hunt

    Edmund B. Nightingale, Orion Hodson, Ross McIlroy, Chris Hawblitzel, and Galen Hunt. Helios: heterogeneous multiprocessing with satel- lite kernels. InProceedings of the ACM SIGOPS 22nd Symposium on Operating Systems Principles, SOSP ’09, 2009

    2009

  41. [41]

    Narappa, and K

    Federico Parola, Sisu Qi, Akhilesh B. Narappa, and K. K. Ramakrishnan. Sure: Secure unikernels make serverless computing rapid and efficient. InProceedings of the 2024 ACM Symposium on Cloud Computing, SoCC ’24, 2024

    2024

  42. [42]

    Inter-VM Shared Memory device.https://www.qemu.org/ docs/master/system/devices/ivshmem.html, 2026

    QEMU. Inter-VM Shared Memory device.https://www.qemu.org/ docs/master/system/devices/ivshmem.html, 2026

    2026

  43. [43]

    Dirigent: Lightweight serverless orchestration

    Lazar Sahraei, Dmitrii Ustiugov, Mahdi Mohammadi Amiri, and An- toni Wolnikowski. Dirigent: Lightweight serverless orchestration. In Proceedings of the ACM SIGOPS 30th Symposium on Operating Systems Principles, SOSP, 2024

    2024

  44. [44]

    Sartakov, Lluís Vilanova, Munir Geden, David Eyers, Takahiro Shinagawa, and Peter Pietzuch

    Vasily A. Sartakov, Lluís Vilanova, Munir Geden, David Eyers, Takahiro Shinagawa, and Peter Pietzuch. ORC: Increasing cloud memory density via object reuse with capabilities. In17th USENIX Symposium on Operating Systems Design and Implementation, OSDI ’23, 2023

    2023

  45. [45]

    Sartakov, Lluís Vilanova, and Peter Pietzuch

    Vasily A. Sartakov, Lluís Vilanova, and Peter Pietzuch. Cubicleos: a library os with software componentisation for practical isolation. In Proceedings of the 26th ACM International Conference on Architectural Support for Programming Languages and Operating Systems, ASPLOS ’21, 2021

    2021

  46. [46]

    Serverless in the wild: Character- izing and optimizing the serverless workload at a large cloud provider

    Mohammad Shahrad, Rodrigo Fonseca, Íñigo Goiri, Gohar Chaudhry, Paul Batum, Jason Cooke, Eduardo Laureano, Colby Tresness, Mark Russinovich, and Ricardo Bianchini. Serverless in the wild: Character- izing and optimizing the serverless workload at a large cloud provider. InProceedings of the 2020 USENIX Annual Technical Conference, ATC ’20, 2020

    2020

  47. [47]

    Legoos: a disseminated, distributed os for hardware resource disaggregation

    Yizhou Shan, Yutong Huang, Yilun Chen, and Yiying Zhang. Legoos: a disseminated, distributed os for hardware resource disaggregation. InProceedings of the 13th USENIX Conference on Operating Systems Design and Implementation, OSDI’18, 2018

    2018

  48. [48]

    Faasm: Lightweight isolation for efficient stateful serverless computing

    Simon Shillaker and Peter Pietzuch. Faasm: Lightweight isolation for efficient stateful serverless computing. InUSENIX ATC, 2020

    2020

  49. [49]

    Frans Kaashoek

    Ariel Szekely, Adam Belay, Robert Morris, and M. Frans Kaashoek. Unifying serverless and microservice workloads with sigmaos. In Proceedings of the ACM SIGOPS 30th Symposium on Operating Systems Principles, SOSP ’24, 2024

    2024

  50. [50]

    Benchmarking, analysis, and optimization of serverless function snapshots

    Dmitrii Ustiugov, Plamen Petrov, Marios Katebzadeh, Michal Sherr, and Boris Grot. Benchmarking, analysis, and optimization of serverless function snapshots. InProceedings of the Twenty-Sixth International Conference on Architectural Support for Programming Languages and Operating Systems, ASPLOS ’21, 2021

    2021

  51. [51]

    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). In28th USENIX Security Symposium, USENIX Security 19, 2019

    2019

  52. [52]

    Slashing the disaggregation tax in heterogeneous data cen- ters with fractos

    Lluís Vilanova, Lina Maudlej, Shai Bergman, Till Miemietz, Matthias Hille, Nils Asmussen, Michael Roitzsch, Hermann Härtig, and Mark Silberstein. Slashing the disaggregation tax in heterogeneous data cen- ters with fractos. InProceedings of the Seventeenth European Conference on Computer Systems, EuroSys ’22, 2022

    2022

  53. [53]

    Peeking behind the curtains of serverless platforms

    Liang Wang, Mengyuan Li, Yinqian Zhang, Thomas Ristenpart, and Michael Swift. Peeking behind the curtains of serverless platforms. In 2018 USENIX Annual Technical Conference, ATC’18, 2018

    2018

  54. [54]

    Isolating functions at the Hardware Limit with Virtines

    Nicholas C Wanninger, Joshua J Bowden, Kirtankumar Shetty, Ayush Garg, and Kyle C Hale. Isolating functions at the Hardware Limit with Virtines. InEuroSys, 2022

    2022

  55. [55]

    WebAssembly System Interface.https://wasi.dev/, 2026

    WASI. WebAssembly System Interface.https://wasi.dev/, 2026

    2026

  56. [56]

    Security implications of memory deduplication in a virtualized environment

    Jidong Xiao, Zhang Xu, Hai Huang, and Haining Wang. Security implications of memory deduplication in a virtualized environment. In2013 43rd Annual IEEE/IFIP International Conference on Dependable Systems and Networks, DSN ’13, 2013

    2013

  57. [57]

    The true cost of containing: A gVisor case study

    Ethan G Young, Pengfei Zhu, Tyler Caraza-Harter, Andrea C Arpaci- Dusseau, and Remzi H Arpaci-Dusseau. The true cost of containing: A gVisor case study. InHotCloud, 2019

    2019

  58. [58]

    Navarro Leija, Ashlie Martinez, Jing Liu, Anna Korn- feld Simpson, Sujay Jayakar, Pedro Henrique Penna, Max Demoulin, Piali Choudhury, and Anirudh Badam

    Irene Zhang, Amanda Raybuck, Pratyush Patel, Kirk Olynyk, Jacob Nelson, Omar S. Navarro Leija, Ashlie Martinez, Jing Liu, Anna Korn- feld Simpson, Sujay Jayakar, Pedro Henrique Penna, Max Demoulin, Piali Choudhury, and Anirudh Badam. The demikernel datapath os architecture for microsecond-scale datacenter systems. InProceedings of the ACM SIGOPS 28th Symp...

    2021

  59. [59]

    Characteri- zation and reclamation of frozen garbage in managed faas workloads

    Ziming Zhao, Mingyu Wu, Haibo Chen, and Binyu Zang. Characteri- zation and reclamation of frozen garbage in managed faas workloads. InProceedings of the Nineteenth European Conference on Computer Systems, EuroSys ’24, 2024. 14

    2024