Middleware 2020

Artifact Evaluation

2021-08-23T14:55:56+02:00

Artifacts Available

Paper #15: PASTRAMI: Privacy-preserving, Auditable, Scalable & Trustworthy Auctions for Multiple Items
Authors: M. Król, A. Sonnino, A. Tasiopoulos, I. Psaras, E. Rivière
Link: https://github.com/harnen/FilecoinPricingMechanism

Paper #51: Fast Training of Deep Learning Models over Multiple GPUs
Authors: X. Yi, Z. Luo, C. Meng, M. Wang, G. Long, C. Wu, J. Yang, W. Lin
Link: https://github.com/eval-submissions/FastT

Paper #115: An OpenMP Runtime for Transparent Work Sharing Across Cache-Incoherent Heterogeneous Nodes
Authors: R. Lyerly, C. Min, C. Rossbach, B. Ravindran
Link: https://github.com/ssrg-vt/popcorn-compiler/tree/master/lib/libopenpop

Paper #127: Prebaking Functions to Warm the Serverless Cold Start
Authors: P. Silva, D. Fireman, T. Pereira
Link: https://github.com/paulofelipefeitosa/prebaking-middleware-2020

Paper #129: Fifer: Tackling Resource Underutilization in the Serverless Era
Authors: J. Gunasekaran, P. Thinakaran, N. Nachiappan, M. Kandemir, C. Das
Link: https://github.com/jashwantraj92/brigade

Paper #150: FLeet: Online Federated Learning via Staleness Awareness and Performance Prediction
Authors: G. Damaskinos, R. Guerraoui, A. Kermarrec, V. Nitu, R. Patra, F. Taiani
Link: https://doi.org/10.5281/zenodo.4048244

Paper #167: Coral-Pie: A Geo-Distributed Edge-compute Solution for Space-Time Vehicle Tracking
Authors: Z. Xu, H. Shah, K. Ramachandran
Link: https://github.com/xzdandy/Coral_Pie

Paper #170: FeGAN: Scaling Distributed GANs
Authors: R. Guerraoui, A. Guirguis, A. Kermarrec, E. Le Merrer
Link: http://doi.org/10.5281/zenodo.4063601

Paper #196: Xanadu: Mitigating cascading cold starts in serverless function chain deployments
Authors: N. Daw, U. Bellur, P. Kulkarni
Link: https://git.cse.iitb.ac.in/synerg/xanadu

Paper #197: Improving the Restore Performance via Physical-Locality Middleware for Backup Systems
Authors: P. Li, Y. Hua, Q. Cao, M. Zhang
Link: https://github.com/iotlpf/HiDeStore

Paper #209: PipeTune: Pipeline Parallelism of Hyper and System Parameters Tuning for Deep Learning Clusters
Authors: I. Rocha, N. Morris, R. Birke, L. Chen, P. Felber, V. Schiavoni
Link: https://github.com/isabellyrocha/pipetune

Paper #238: Consistent and Secure Network Updates Made Practical
Authors: J. Lembke, S. Ravi, P. Roman, P. Eugster
Link: https://gitlab.com/robust-sdn/cicero/

Paper #295: SplitServe: Efficiently Splitting Apache Spark Jobs Across FaaS and IaaS
Authors: A. Jain, A. Baarzi, N. Alfares, G. Kesidis, B. Urgaonkar, M. Kandemir
Link: https://github.com/PSU-Cloud/splitserve

Paper #425: Cost-Effective Data Feeds to Blockchains via Workload-Adaptive Data Replication
Authors: K. Li, Y. Tang, J. Chen, Z. Yuan, C. Xu, J. Xu
Link: https://github.com/syracuse-fullstacksecurity/GRuB

Artifacts Evaluated - Functional

Paper #15: PASTRAMI: Privacy-preserving, Auditable, Scalable & Trustworthy Auctions for Multiple Items
Authors: M. Król, A. Sonnino, A. Tasiopoulos, I. Psaras, E. Rivière
Link: https://github.com/harnen/FilecoinPricingMechanism

Paper #115: An OpenMP Runtime for Transparent Work Sharing Across Cache-Incoherent Heterogeneous Nodes
Authors: R. Lyerly, C. Min, C. Rossbach, B. Ravindran
Link: https://github.com/ssrg-vt/popcorn-compiler/tree/master/lib/libopenpop

Paper #127: Prebaking Functions to Warm the Serverless Cold Start
Authors: P. Silva, D. Fireman, T. Pereira
Link: https://github.com/paulofelipefeitosa/prebaking-middleware-2020

Paper #150: FLeet: Online Federated Learning via Staleness Awareness and Performance Prediction
Authors: G. Damaskinos, R. Guerraoui, A. Kermarrec, V. Nitu, R. Patra, F. Taiani
Link: https://doi.org/10.5281/zenodo.4048244

Paper #170: FeGAN: Scaling Distributed GANs
Authors: R. Guerraoui, A. Guirguis, A. Kermarrec, E. Le Merrer
Link: http://doi.org/10.5281/zenodo.4063601

Paper #197: Improving the Restore Performance via Physical-Locality Middleware for Backup Systems
Authors: P. Li, Y. Hua, Q. Cao, M. Zhang
Link: https://github.com/iotlpf/HiDeStore

Paper #209: PipeTune: Pipeline Parallelism of Hyper and System Parameters Tuning for Deep Learning Clusters
Authors: I. Rocha, N. Morris, R. Birke, L. Chen, P. Felber, V. Schiavoni
Link: https://github.com/isabellyrocha/pipetune

Paper #425: Cost-Effective Data Feeds to Blockchains via Workload-Adaptive Data Replication
Authors: K. Li, Y. Tang, J. Chen, Z. Yuan, C. Xu, J. Xu
Link: https://github.com/syracuse-fullstacksecurity/GRuB

Student Volunteer Instructions

2020-12-03T02:02:39+01:00

Workshop Instructions

Your role is to open the Zoom channel and assist the workshop organizers, depending on what they need. You need to find that out.

Before the workshop (do it today!):

Request access to the spreadsheet by following the link while logged into your Google account and then click the request access button.
Please reach out to the workshop organizers through email and check if they have any specific instructions for you.
Check with them if they have collected the video files and will play them or if they are expecting you to play the video files.
Make sure that you have the most recent version of the Zoom client installed on your local machine.

At the day of the workshop:

Make sure that you are logged into Slack.
Open the Zoom channel at least 30 minutes before the scheduled start time, in the following way:
Log into zoom.us (make sure that you are using the account that corresponds to your channel), go to scheduled meetings, and start the meeting.
The account name is acm@conference.acm.org, password will be sent out through email
The meeting should start using your local client.
Look out for the organizers to join and make them co-host. This is important so that the session does not drop in case you would get disconnected. They are instructed to join earlier than the start date.
Make sure that the session is recorded. On Monday, we are trying if we can record all sessions to the cloud. The scheduled sessions should record automatically once started.

Check with the organizers if they can run the channels by themselves or if they need help, e.g., with reading questions asked on Slack.

Main Conference Instructions

Your role is to play the videos of the session that you are scheduled for.

Before the conference (do it today!):

Request access to the spreadsheet by following the link while logged into your Google account and then click the request access button.
Make sure that you have the most recent version of the Zoom client installed on your local machine.
Check the video files of your session, which you find in the spreadsheet.
Make sure that they play flawlessly on your machine and that the audio is good.
Check that the length of the video is around 10 minutes or less.
Check if the video has subtitles, if so, mark this information for yourself so that you remember enabling them when playing

On the days of the conference:

The meeting will be already open because we are using the same meeting for all sessions.
Make sure that you join at least 15 minutes before the start of the session and make yourself known to me by message (we might have 100+ people in the session)
I will assign the co-host role to you
You will start sharing your screen about 10 minutes before the start of the session.
Make sure that you check the options “share computer audio” and “optimize screen share for video clip”

You should have only one browser open on the screen that you share but all videos already open in tabs so that switching is efficient
On the first tab, show the 2020.middleware-conference.org/ homepage until the session starts
The session chair starts the session, start the first presentation when asked to
There is a Q&A after every talk. Check the Slack channel for written questions and announce them should the session chair miss them.
Start the next presentation when the session chair indicates that no more questions can be taken.

If you have any questions, please contact me.

Keynote: Developing and Assessing Middleware and Applications for Smart Cities

2020-11-25T00:38:44+01:00

Smart Cities combine advances in the Internet of Things, Big Data, Social Networks, Cloud Computing, and Data Science technologies with the demand for cyber-physical applications in areas of public interest, such as Health, Public Safety, and Mobility. The end goal is to leverage the use of city resources to improve the quality of life of its citizens. Achieving this goal, however, requires advanced support for the development and operation of applications in a complex and dynamic environment. Middleware platforms can provide an integrated infrastructure that enables solutions for smart cities by combining heterogeneous city devices and providing unified, high-level facilities for the development of applications and services. Although several smart city platforms have been proposed in the literature, there are still open research and development challenges related to their scalability, maintainability, interoperability, and reuse in the context of different cities, to name a few. Unfortunately, available platforms lack extensive scientific validation. To contribute in this direction, we are working on InterSCity, a microservices-based, open-source, smart city platform that works as a middleware enabling the collaborative development of large-scale systems, applications, and services. In this talk, we present the architecture of the InterSCity platform, show examples of applications developed around it, and discuss how to assess the performance of such a platform with large-scale workloads produced via simulation based on realistic data.

About the speaker

Fabio Kon is a Full Professor of Computer Science at the University of São Paulo. His research interests include Distributed, Scalable Systems, Software Engineering, Open Source, and Data Science. In recent years, Fabio has focused his research on Smart Cities (https://interscity.org), Evidence-based Public Policymaking, and DevOps. Fabio is an ACM Distinguished Scientist and wishes he had more time to play his vibraphone.

Keynote: Teaching Distributed Systems in the Digital Age

2020-11-25T00:38:19+01:00

Having taught Distributed Systems for many years and confronted with having been asked to take over the course from a colleague who was retiring, I felt it was time for a change. Distributed systems have a very solid and stable basis, theoretically as well as practically (considering the many systems that have been in use for often decades). There is simply no real need to teach students what they can pick up from any reasonable textbook. The challenge in teaching Distributed Systems is to let students understand the inherent complexities: where does performance come from, where can I find the security measures, why does it hurt here when some part fails over there, and so on.

In this, anticipated highly interactive keynote, I'd like to discuss a completely different setup that is based on just a few principles: flip the classroom, good explanations come from thorough understanding, and understanding one problem very well is enough to open doors to understanding other issues in Distributed Systems. Rather than letting students dig into well-known problems, I believe it is extremely insightful to let them concentrate on controversial issues. Controversial because statements are made (in blogs and postings and scientific articles) about certain solutions that should be seriously questioned. Oftentimes, those statements are false or inaccurate, but it requires effort to discover why.

About the speaker

Maarten van Steen has been full professor in Large-Scale Distributed Systems since 2002 and has published on a wide variety of subjects related to distributed systems. Together with Andrew Tanenbaum he has authored a reference textbook on the matter that has seen its way to many universities worldwide. He is currently Scientific Director of the Digital Society Institute of the University of Twente, in essence a virtual institute that bundles all the university's research on digitalization. Before joining Twente in 2015, he was member of the Computer Systems group at the Vrije Universiteit Amsterdam for some 20 years.

Working in the field of distributed systems ever since 1983, he continues to struggle understanding many of its aspects, especially when scalability is at stake. It is has driven him to believe that only the truly simple solutions will work and that complex solutions (including some of his own) come from a lack of understanding the essentials.

Program - Doctoral Symposium

2020-11-18T10:28:03+01:00

Tuesday, 8th December

all times in CET

8:50 - 9:00	Welcome & Introduction
9:00 - 9:35	Session 1 Session chairs: Animesh Trivedi, Arpan Gujarati, and Pramod Bhatotia Keynote: Lalith Suresh (VMWare Research) https://lalith.in/2020/09/27/Low-Level-Advice-For-Systems-Research/
9:35 - 9:45	break
9:45 - 10:20	Session 2 (5 min per talk, ~3 min for questions) Session chairs: Paris Carbone, Aastha Mehta, and Pramod Bhatotia (1) C. Berger - "Towards Environmental-Adaptive and Performance-Resilient Consensus in Distributed Ledger Technology" (2) G. Russo - "Model-based auto-scaling of distributed data stream processing applications" (3) V. Podolskiy - "Rubberband: Enabling Elastic Federated Learning with the Temporary State Management" (4) M. Kamalian - "Fog Transport Mode Detection"
10:30 - 11:05	Session 3 (5 min per talk, ~3 min for questions) Session chairs: Thomas Pasquier, Antonio Barbalace, and Ana Varbanescu (5) G. Frascaria - "Griffin: A Storage Service for Granular Edge Applications" (6) S. Tavakolisomeh - "Selecting a JVM Garbage Collector for Big Data and Cloud Services" (7) P. Eichhammer - "Towards Autonomous Security Monitoring in Internet-of-Things Infrastructures" (8) J. Gunasekaran - "Minimizing Cost and Maximizing Performance for Cloud Platforms"
11:05 - 11:15	break
11:15 - 11:50	Session 4 (5 min per talk, ~3 min for questions) Session chairs: Etienne Riviere, Anjo Vahldiek-Oberwagner, and Ana Varbanescu ( 9) A. Isstaif - "Self-managed services using MirageOS unikernels" (10) A. Khordadi - "Execution Migration of Managed Languages Programs Among Heterogeneous-ISA Processing Units" (11) I. Zavalyshyn - "Building Private-by-Design IoT Systems" (12) P. Schmiedmayer - "Apodini: An Internal Domain Specific Language to Design Web Services"
11:50 - 12:00	Closing Remarks

Video Submission Instructions

2020-11-06T03:21:49+01:00

Video Submission Deadline: November 27th, 2020

Uploading the Video to YouTube

Sign in to YouTube.
At the top right corner, select the Create video or more button -> Upload Video
Select the video file you would like to upload.
While your video is being processed, click on “Details” of the video to fill in information. Your video title should be in the following format: TITLE (ACM MIDDLEWARE 2020). Note that for tracks where this is applicable (e.g., industry track), the title should contain the prefix.
In the description section, please include the track (Research, Industry, ...), paper ID (which can be found from HotCRP) and the names of the authors.
Under visibility, please select “Unlisted” or “Public”. The difference is that an unlisted video will not be indexed by the YouTube search engine. Click “Save” once done.
You can optionally go to the “Subtitles” section to manually enter subtitles for your video, which will be displayed during the live broadcast. Please ensure the subtitles are accurate.
Once your video is available, please copy the link by viewing the video page on YouTube and sharing its link.

Emailing the Youtube Link

Email your link to middleware2020@tudelft.nl

The email subject must be titled with your paper ID # and presenter last name:

MIDDLEWARE 2020 Video Link: PAPERID#_PRESENTER-LASTNAME

The email body should contain:

Paper ID and track
YouTube video link
Name of the speaker in the video
A short 3-4 lines biography of the speaker

If you have any difficulty following these instructions, please contact the chairs of your track or the general chairs immediately.

Presentation Instructions

2020-11-06T11:00:54+01:00

Submission deadline for presentation videos: November 27, 2020

To ensure smooth operation for our online program, talks are pre-recorded and uploaded to be streamed during each session. Authors must attend their respective Zoom sessions for the Q&A immediately following the broadcast of their video, according to the moderation of the session chair.

The presentation time is 10 minutes for research and industry track papers. Videos which significantly exceed this time limit can be cut off by the moderators during the presentation. The files have to be uploaded to YouTube and the video quality has to be at least HD (720p).

Following every presentation, there is a Q&A session of 5 minutes which the author is required to attend. At the end of the session, there is a panel discussion comprising all presenters of the session.

Authors should treat the video recording as a direct substitute for a physical presentation. The audio should contain a spoken narration of the presentation. The accompanying video should contain a full walkthrough of the presentation slides, which are clearly visible. In addition, the presenter is strongly encouraged to overlay a thumbnail-sized camera feed showing the speaker. In addition, please add some 2-3 discussion points to your conclusion slide. As the talks are intentionally kept short for the online format, it is likely that there are points not discussed during the presentation, such as technical details or future work. Please include these in the conclusion slide to facilitate questions during the Q&A.

Uploading the Video to YouTube

Sign in to YouTube.
At the top right corner, select the Create video or more button -> Upload Video
Select the video file you would like to upload.
While your video is being processed, click on “Details” of the video to fill in information. Your video title should be in the following format: TITLE (ACM/IFIP MIDDLEWARE 2020). Note that for tracks where this is applicable (e.g., industry track), the title should contain the prefix.
In the description section, please include the track (Research, Industry, ...), paper ID (which can be found from HotCRP) and the names of the authors.
Under visibility, please select “Unlisted” or “Public”. The difference is that an unlisted video will not be indexed by the YouTube search engine. Click “Save” once done.
You can optionally go to the “Subtitles” section to manually enter subtitles for your video, which will be displayed during the live broadcast. Please ensure the subtitles are accurate.
Once your video is available, please copy the link by viewing the video page on YouTube and sharing its link.

Emailing the Youtube Link

Email your link to middleware2020@tudelft.nl

The email subject must be titled with your paper ID # and presenter last name:

MIDDLEWARE 2020 Video Link: PAPERID#_PRESENTER-LASTNAME

The email body should contain:

Paper ID and track
YouTube video link
Name of the speaker in the video
A short 3-4 lines biography of the speaker

If you have any difficulty following these instructions, please contact the chairs of your track or the general chairs immediately.

Conference Program

2020-12-09T23:21:32+01:00

Detailed Program

Monday Dec. 7th

M4IoT Workshop

DIDL Workshop

Containers Workshop

Tuesday Dec. 8th

Wednesday Dec. 9th

10:00-11:30	Opening and Keynote (Session Chairs: Jan S. Rellermeyer and Rüdiger Kapitza)
	Opening Middleware 2020
	Keynote Maarten van Steen

13:00-14:30	Research Track: Session 1 (Session Chair: Hans P. Reiser)
	Prebaking Functions to Warm the Serverless Cold Start (Presenter: Paulo Silva)
	Resilient Cloud-based Replication with Low Latency (Presenter: Michael Eischer)
	Practical Active Revocation (Presenter: Stefan Contiu)
	secureTF: A Secure TensorFlow Framework (Presenter: Do Le Quoc)
	On Delivery Guarantees in Distributed Content-Based Publish/Subscribe Systems (Presenter: Pooya Salehi)

15:30-17:00	Industry Track: Session (Chairs: Ioana Giurgiu and Sameh Elnikety)
	mmFilter: Language-Guided Video Analytics at the Edge (Presenter: Zhiming Hu)
	blockNDP: Block-storage Near Data Processing (Presenter: Martin Decky)
	DiagSys: network and third-party web-service monitoring from the browser’s perspective (Presenter: Loïck Bonniot)
	HopsFS-S3: Extending Object Stores with POSIX-like Semantics and more (Presenter: Mahmoud Ismail)
	Primula: a Practical Shuffle/Sort Operator for Serverless Computing (Presenter: Germán T. Eizaguirre)
	Secure Volume Hot-plugging for Containers (Presenter: Eunsoo Evan Park)

18:00-19:30	Posters and Demos (Chairs: Angela Nicoara and Ihor Kuz)
	Demo: Hierarchical Topic Aggregation for Geospatial Applications through Processing Extensions in Message Brokers (Presenter: Josef Spillner)
	Poster: Version Detection for Software Discovery in the Cloud (Presenter: Sadie Allen)
	Poster: DynPubSub: A Peer To Peer Overlay For Topic-Based Pub/Sub Systems Deployed at the Edge (Presenter: Chamseddine Bouallegue)
	Demo: Given 2n Eyeballs, All Quality Flaws Are Shallow (Presenter: Josef Spillner)
	Demo: Practical Trade-Offs in Integrity Protection for Binaries via Ethereum (Presenter: Oliver Stengele)

20:00-21:00	Virtual Reception

Thursday Dec. 10th

10:00-11:30	Research Track: Session 2 (Session Chair: Mina Sedaghat)
	MATCH: A Decentralized Middleware for Fair Matchmaking In Peer-to-Peer Markets (Presenter: Martijn de Vos)
	PipeTune: Pipeline Parallelism of Hyper and System Parameters Tuning for Deep Learning Clusters (Presenter: Isabelly Rocha)
	Fast Training of Deep Learning Models over Multiple GPUs (Presenter: Xiaodong Yi)
	BoLT: Barrier-optimized LSM-Tree (Presenter: Dongui Kim)
	JellyFish: A Fast Skip List with MVCC (Presenter: Leeju Kim)

13:00-14:30	Research Track: Session 3 (Session Chair: Vincenzo Massimiliano Gulisano)
	Consistent and Secure Network Updates Made Practical (Presenter: James Lembke)
	FLeet: Online Federated Learning via Staleness Awareness and Performance Prediction (Presenter: Georgios Damaskinos)
	TEEMon: A continuous performance monitoring framework for TEEs (Presenter: Robert Krahn)
	FeGAN: Scaling Distributed GANs (Presenter: Arsany Guirguis)
	Poseidon: Safe, Fast and Scalable Persistent Memory Allocator (Presenter: Madhava Krishnan)

15:30-17:00	Research Track: Session 4 (Session Chair: Roy Campbell)
	WASP: Wide-area Adaptive Stream Processing (Presenter: Abhishek Chandra)
	SplitServe: Efficiently Splitting Apache Spark Jobs Across FaaS and IaaS (Presenter: Aman Jain)
	SR3: Customizable Recovery for Stateful Stream Processing Systems (Presenter: Pinchao Liu)
	Sledge: a Serverless-first, Light-weight Wasm Runtime for the Edge (Presenter: Gabriel Parmer)
	Tackling Resource Underutilization in the Serverless Era (Presenter: Jashwant Raj Gunasekaran)

18:00-19:30	Keynote and Test of Time Award (Session Chairs: Dilma da Silva and Nalini Venkatasubramanian)
	Keynote Fabio Kon
	Test of Time Award Ceremony

20:00-21:00	Open Business Meeting (Session Chair: Hans-Arno Jacobsen)
	Middleware 2021 Address

Friday Dec. 11th

10:00-11:30	Research Track: Session 5 (Session Chair: Valerio Schiavoni)
	PASTRAMI: Privacy-preserving, Auditable, Scalable & Trustworthy Auctions for Multiple Items (Presenter: Michał Król)
	A practical approach for updating an integrity-enforced operating system (Presenter: Wojciech Ozga)
	Regaining Lost Seconds: Efficient Page Preloading for SGX Enclaves (Presenter: Lizhi Wang)
	Improving the Restore Performance via Physical-Locality Middleware for Backup Systems (Presenter: Pengfei Li)
	Xanadu: Mitigating cascading cold starts in serverless function chain deployments (Presenter: Nilanjan Daw)

13:00-14:30	Research Track: Session 6 (Session Chair: Francois Taiani)
	Cost-Effective Data Feeds to Blockchains via Workload-Adaptive Data Replication (Presenter: Yuzhe Tang)
	Mitigating Stragglers in the Decentralized Training on Heterogeneous Clusters (Presenter: Donglin Yang)
	Coral-Pie: A Geo-Distributed Edge-compute Solution for Space-Time Vehicle Tracking (Presenter: Zhuangdi (Andy) Xu)
	An OpenMP Runtime for Transparent Work Sharing Across Cache-Incoherent Heterogeneous Nodes (Presenter: Robert Lyerly)
	DoubleFaceAD: A New Datastore Driver Architecture to Optimize Fanout Query Performance (Presenter: Shungeng Zhang)
14:30	Closing Remarks