Concurrent Systems (prof. Gorla) - AA 2021-2022

The class is focused on the foundational aspects and on the formal/mathematical semantics of concurrent systems. The class is structured in two main parts. The first part describes the main characteristics and the basilar problems of every concurrent system (mutual exclusion, synchronization, atomicity, deadlock/livelock/starvation, ...) and the relative solutions at the implementation level (semaphores, monitors, system primitives, ...). Furthermore, more evolute notions are shown, like: failure detectors, their implementation and their use to obtain wait-free implementations; universal object, consensus object and consensus number; transactional memory, ... The second part of the course describes the preliminary notions of a minimal concurrent language called CCS (execution of parallel processes through labelled transition systems, interleaving semantics, syntonization, non-determinism, process simulability) and presents a mathematical model, with different features for the specification and the analysis of systems written in such a language.

In the time left, we shall have lectures in the form of seminars where more advanced programming mechanisms (like name creation and exchange, type systems for the verification of properties, cryptography, distribution, truly concurrent semantics) will be presented.

The course integrates didactic parts to recent research problems.

Timetable

2nd semester (end of February --> end of May).

Text books

First part:

• M. Raynal: Concurrent Programming: Algorithms, Principles and Foundations. Springer, 2013. (chapters 1, 2, 3, 4, 5, 10, 14, 16 and part of 17).

Second part:

• R. Milner. Communicating and Mobile Systems. Cambridge University Press, 1999. (chapters: 1, 2 (no 2.3), 3, 4 (no 4.1, 4.4, 4.5 and 4.6), 5 (no prop. 5.2, ex 5.3, lemma 5.4/5.5, theor. 5.6, def. 5.17, prop. 5.18, prop. 5.23), 6 (no 6.3) and 7 (only 7.2 and 7.3).
• R. Cleaveland and S. Smolka. Process Algebra. In Encyclopedia of Electrical Engineering, John Wiley & Sons, 1999. Available at http://www.cs.umd.edu/~rance/publications/papers/ee99.ps.gz. (sect. 3.2, 4.1 and 4.2).
• R. Cleaveland and O. Sokolsky. Equivalence and Preorder Checking for Finite-State Systems. In "Handbook of Process Algebra," pp. 391-424, Elsevier, 2001. Available at http://www.cis.upenn.edu/~sokolsky/PAhandbook.pdf. (sect. 3.1).

Third part:

• J. Parrow. An Introduction to the pi-Calculus. In Handbook of Process Algebra, pages 479-543. Elsevier, 2001. Available at http://user.it.uu.se/~joachim/intro.ps
• R. Milner. The Polyadic pi-Calculus: A Tutorial. In Logic and Algebra of Specification. Springer-Verlag, 1993. Available at http://www.lfcs.inf.ed.ac.uk/reports/91/ECS-LFCS-91-180/ECS-LFCS-91-180.ps
• D. Sangiorgi, D. Walker. The Pi-Calculus - A Theory of Mobile Processes. Cambridge University Press, 2001. The reference book on the pi-calculus; maybe, quite difficult for a master student.
• http://www.cs.unibo.it/~zavattar/Publication/icalp03.ps. Comparison of recursion and replication in process calculi.
• http://www.cs.unibo.it/~sangio/DOC_public/pi.ps.gz. The classical notion of subtyping in the pi-calculus.
• http://www.cs.unibo.it/~zavattar/Publication/tcs00.ps. A comparison of three different kinds of output actions in a CCS-like language.
• http://lucacardelli.name/Papers/Mobility%20and%20Security.A4.pdf. A nice survey on a calculus very trend a few years ago. There are two logically distinct parts: I+II+III and IV+V.
• http://lucacardelli.name/Papers/Brane%20Calculi.pdf. A funny work on a formalism for bio-inspired computing.
• https://www.cs.bham.ac.uk/~mdr/research/papers/pdf/11-applied-pi.extended.pdf. A survey on one of the best known cryptographic versions of the pi-calculus.

Detailed Program

The course is split into 3 parts:

• The first part studies foundational problems of concurrent systems:
  • sequential vs concurrent programs
  • Process synchronization (competition vs cooperation)
  • Safety and liveness properties; a hierarchy of liveness properties (deadlock freedom, starvation freedom, bounded bypass)
  • Mutual exclusion: atomic registers (algorithms by Peterson and Lamport; how to obtain a starvation free lock object from a deadlock free lock object), with specialized hardware primitives (test&set, compare&swap, fetch&add) and with non-atomic registers (Bakery algorithm).
  • Semaphores, monitors and their use for solving classical problems (producers-consumers, readers-writers, rendezvous, dining philosophers)
  • Transactional memory
  • Atomicity and its properties
  • mutex-free concurrency: liveness revisited (obstruction freedom, non-blocking and wait-freedom); mutex-free implementation of an unbounded stack; failure detectors Omega_X and Diamond_P, and the associated construction of non-blocking and wait-free implementations from an obstruction free implementation; hints on how to implement the Omega failure detector.
  • Universal object; consensus object and its universality; primitives hierarchy based on the consensus number.

• The second part is focused on CCS (Calculus of Communicating Systems [Milner:1980]):
  • non-determinism
  • Labeled transition systems
  • recursion
  • parallel composition (interleaving semantics)
  • synchronization
  • restriction
  • bisimulation: strong vs weak, axiomatization, logical characterization, algorithmic verification.

• The third part will present possible enhancements of the previous material (every lecture will be on a different topic). By following the students' interests, some of the following topics can be presented:
  • communication: name creation and passage
  • polyadic communication (type system for correct communications and encodability in the base calculus)
  • asynchronous communication and encodability of synchrony
  • higher-order communication and its encodability in the base calculus
  • encoding of the lambda-calcolus
  • a model for object-oriented languages
  • concurrent calculi with cryptography or distribution
  • truly concurrent semantics

Diary

Here, [R] denotes Raynal's book; [CCS] and [PI] denote teacher's lecture notes on CCS and pi-calculus (that are a unified and coherent presentation of the texts for the second and the third part of the course listed above).

Feb 23rd, 2022 ([R]: chapt.1, excluding sect.1.2.5; 2.1.1, 2.1.2, 2.1.3. Lecture Notes): Sequential vs Multiprocess Program; process synchronization (competition and cooperation); the mutual exclusion problem; safety and liveness properties; a hierarchy of liveness properties (bounded bypass; starvation freedom; deadlock freedom). Atomic read/write registers; mutex for 2 processes (Peterson algorithm).

Feb 25th, 2022 ([R]: 2.1.4; 2.1.5 - only the idea, no algo, no proofs; 2.1.6; 2.1.7. Lecture Notes): Generalizing Peterson's algorithm to n processes. Algorithms with better performances when no concurrency is present: with a tournament tree of 2-processes competitions (O(log n)); Lamport's fast mutex algorithm (with constant time when there is no contention).

March 2nd, 2022 ([R]: 2.2 and 5.2.3; 2.3.1, 2.3.2. Lecture Notes - Lecture Notes): From Deadlock freedom to Starvation freedom using atomic r/w-registers; mutex with specialized HW primitives (test&set; swap; compare&set; fetch&add); the ABA problem with the compare&set. Safe registers: Lamport's Bakery algorithm

March 4th, 2022 ([R]: 2.3.3. Lecture Notes): Aravind's bounded algorithm.

March 9th, 2022 ([R]: 3.1, 3.2.1, 3.2.2, 3.2.4. Lecture Notes): Concurrent objects. Semaphores and their implementation. Use of semaphores in the producer/consumer problem (both for single producer/consumer and for multiple ones) and in the readers/writers problem (both with weak/strong priority to the readers and with weak priority to the writers).

March 11th, 2022 ([R]: 3.3.1, 3.3.2, 3.3.4, 3.3.5. Lecture Notes1, Lecture Notes2): Monitors: concept and implementation through semaphores; implementing rendez-vous through monitors. Readers/writers through monitors. The problem of the "Dining Philosophers": solutions that break symmetry (through semaphores) and a symmetric solution (through monitors).

March 16th, 2022 : First classwork.

March 18th, 2022 ([R]: chap. 4. Lecture Notes): Atomicity: formal definition and compositionality; possible variants and their non-composability.

March 23rd, 2022 ([R]: 10.1, 10.2, 10.3, 10.5. Lecture Notes): Software Transactional Memory. Opacity and a Logical clock-based STM system (TL2). Virtual World Consistency and a vector clock-based STM system (REMARK: to better understand the difference between virtual world consistency and opacity, look at this paper: https://www.sciencedirect.com/science/article/pii/S0304397512004021/pdf?md5=78dd63551d097b6a3212285669d20ff8&pid=1-s2.0-S0304397512004021-main.pdf, sections 2.6 and 3.2).

March 25th, 2021 ([R]: 5.1, 5.2.1, 5.2.2, 5.2.5, 5.2.6. Lecture Notes): Mutex-free concurrency: problems of mutex and notion of mutex-freedom, progress conditions. Examples: splitter, timestamp generator and stack (based on swap plus fetch&add; based on compare&set); progress conditions for these examples.

March 30th, 2022 ([R]: 5.3.1, 5.3.2, 5.3.3, 5.3.4; 17.7, 17.7.1, 17.7.2 -- only the idea, not the formal protocol. Lecture Notes): From obstruction-freedom to non-blocking through failure detector Omega_X; hints on the implementation of Omega_X. From obstruction-freedom to wait-freedom through failure detector Diamond_P; hints on the implementation of Diamond_P.

April 1st, 2022 ([R]: 14.1, 14.2, 14.3, 14.5. Lecture Notes): Universal object and consensus object; universality of consensus (unbounded wait-free construction).

April 6th, 2021 ([R]: 16.1, 16.2, 16.3, 16.4.1, 16.4.3, 16.4.4, 16.4.5 -- adapted to test&set -- 16.5.1, 16.6. Lecture Notes): Binary vs multivalued consensus. Consensus number (for atomic R/W registers, test&set, swap, fetch&add, compare&swap) and consensus hierarchy.

April 8th, 2022 ([CCS]: chapters 1 and 2. Lecture Notes): Automata for describing process behaviors (notion of LTS); inadequacy of trace equivalence for equating non-deterministic processes; simulation, double simulation and bisimulation. Properties of bisimilarity.

April 13th, 2022 : Second classwork.

April 20th, 2022 ([CCS]: chapters 1 and 2.Lecture Notes and Lecture Notes): Syntax for non-deterministic processes: from LTS to the syntax and vice versa. Examples: counter and queue. Process interaction, parallel composition and name restriction.

April 22nd, 2022 ([CCS]: chapter 3.1, 3.2, 3.3. Lecture Notes): A first proof technique for proving properties of LTSs: the case of image-finiteness. A second proof technique for proving properties of LTSs: the case of closure under substitutions. A simple example that uses bisimilarity for proving an implementation equivalent to its specification.

April 27th, 2022([CCS]: 3.4, 4.1, 4.2. Lecture Notes): Congruence of Bisimilarity. Weak bisimilarity: basic properties, comparison with strong bisimilarity and fundamental laws.

April 29th, 2022([CCS]: 4.3. Lecture Notes): Examples using weak bisimilarity: the factory, the lottery and the scheduler.

May 4th, 2022: Third classwork.

May 6th, 2022 ([CCS]: 5.1. Lecture Notes): An inference system for strong bisimilarity: soundness and completeness for finite processes. The tau laws for weak bisimilarity. Verifying the equivalence of a specification and an implementation through the inference system.

May 11th, 2022 ([CCS]: 5.3. Lecture Notes): The Kennelakis and Smolka Algorithm for bisimulation on finite state LTSs; soundness and complexity.

May 13th, 2022 ([CCS]: 5.2. Lecture Notes): A logical characterization of bisimilarity and its use to show process inequivalences; sub-logics for double simulation and for trace equivalence; a logic for weak bisimilarity.

May 18th, 2022 ([PI]: Chapt.1. Lecture Notes): The pi-calculus: from CCS, through value-passing CCS; syntax and reduction semantics; implementing recursive parametric definitions through replication.

May 20th, 2022 ([PI]: Chapter 2.1, 2.2. Lecture Notes): The polyadic version of the calculus and a type system for correct communications. Encoding the polyadic version of the calculus in the monadic version. The asynchronous calculus and the encoding of synchrony into asynchrony.

May 25th, 2022 ([PI]: Chapt. 2.3, 4. Lecture Notes): The higher-order calculus, encoding replication and encoding HOpi in the first order paradigm. Encoding the lazy lambda-calculus and an object-oriented calculus.

May 27th, 2022: Fourth classwork.

Exam modalities

There are two possible ways to pass the exam:

• First modality (better suited for students who regularly attend the classes): classworks, approx one every two weeks. Students should pass 3 over 4 classworks. There will be 2 classworks for the first part and 2 for the second+third part. The final mark will be the average of the 3 better marks in such classworks
• Second modality: written and oral exam on all the topics presented during the classes.

News

Course Material and Googlegroup Students can have access to all text books and lecture notes by sending an email to the teacher and ask to be included in the googlegroup of the curse. The first posts of this group contain in attachment all what you need for studying. Please, register with the email address that you most frequently access, because important and urgent news about the course will be primarily posted on this group. Furthermore, you'll receive all information for remotely attending the classes on such googlegroup (zoom meeting number and password). Of course, students can also come in presence, by following Sapienza's rules.

-- DanieleGorla - 20 Jan 2016