All lectures will be held in Kaihuset.
Tuesday, August 14, 2018
09:00 – 16:45: City-center information desk
Location: Kaigata 5 (the Sparebank 1 building)
If you have questions, please drop by city center information desk located in Kaigata 5.
15:00: Airport shuttle to city center.
Location: Outside the rental car exit
16:45: Main Transport: Boat Cruise from Tromsø to Sommarøy ?
Location: Prostneset Terminal (City Center)
The boat docks in the center of Tromsø. We aboard the boat through the newly built Terminal in Prostneset, Tromsø.
19:00 – 22:00: Dinner reception at Sommarøy
Location: Sommarøy Arctic Hotel
19:30: Airport shuttle to Sommarøy
Location: Look for the driver in the baggage reclaim area.
The shuttle has limited capacity. Seats are reserved for participants arriving on flights SK4424 and SK4430.
20:30: Airport shuttle to Sommarøy
22:30: Airport shuttle to Sommarøy
Wednesday, August 15, 2018
07:30 – 08:30: Breakfast
08:45 – 09:00: Welcome, Introduction
09:00 – 09:50: Session 1. Distributed Computing – External Simplicity, Internal Complexity
Speaker: Dag Johansen
Numerous elite sports are currently undergoing a silent revolution due to the rapidly-increasing use of quantified data and associated analytics. An abundance of objective performance data from individual players can be collected automatically during practices and competitions, and some data even from their lives outside the sports arenas. Further algorithmic analysis of this data might provide vital insights for individual injury prevention and training personalization, and a foundation for evidence-based decisions for team performance improvements.
This talk gives a brief introduction to distributed systems, before initial experiences from building and deploying heterogeneous distributed systems in the elite soccer domain are presented. We will, particularly, focus on distributed systems we have developed and put to good use aiding and complementing sport-specific human competence, knowledge, and intuition.
09:50 – 10:05: Break
10:05 – 10:55: Session 2. If You’re Not Writing a Program, Don’t Use a Programming Language. ?
Speaker: Leslie Lamport
Algorithms are not programs. They can and should be written with math rather than programming languages or pseudo-languages. This applies to many more algorithms than the ones taught in algorithm courses.
10:55 – 11:10: Break
11:10 – 12:00: Session 3. Lamport’s Turing Award talk ?
Speaker: Leslie Lamport
Those who cannot remember the past may not be condemned to repeat it, but they’ll miss some interesting stories. A little of what they’d miss.
12:00 – 14:00: Lunch
14:00 – 14:50: Session 4. Consistency vs Performance 1
Speaker: Lorenzo Alvisi
A fundamental challenge in the design of distributed data stores is to balance two conflicting goals: strong consistency (which directly translates into ease of programming) and performance. We’ll explore this dilemma across distributed systems and databases–we will understand the why and the how of this tension, and become familiar with notions such as linearizability, eventual consistency, causal consistency, and the ACID/BASE dualism. Then, we will deep dive into recent attempts in both fields at bridging this gap.
14:50 – 15:05: Break
15:05 – 15:55: Session 5. Consistency vs Performance 2
Speaker: Lorenzo Alvisi
16:30 – 20:00: Social program: Hike to Hillesøytoppen (211 moh) and visit Halvard’s Place ?
The hike will take some effort but is doable for most. Bring your walking or hiking shoes. If the hike looks like it will be too much, let us know. We have alternatives.
20:00 – 22:00: Dinner
Thursday, August 16, 2018
07:30 – 08:45: Breakfast
09:00 – 09:50: Session 6. Fail-Stop Replication 1 ?
Speaker: Robbert van Renesse
State Machine Replication protocols such as Paxos have received much theoretical attention lately and many implementations are available, but compared to traditional Primary-Backup (PB) protocols they suffer from much inefficiency. With recent advances in failure detection, PB protocols are still of practical interest, but in their presented forms they do not support self-configuration and recovery from total failure. We show how Chain Replication and Primary-Backup can indeed support self-configuration and recover from total failures.
09:50 – 10:05: Break
10:05 – 10:55: Session 7. Fail-Stop Replication 2 ?
Speaker: Robbert van Renesse
10:55 – 11:10: Break
11:10 – 12:00: Session 8. Byzantine Consensus to Blockchains 1 ?
Speaker: Christian Cachin
Although practical Byzantine consensus protocols have been available for two decades, they have never been deployed in production until very recently. The widespread interest in cryptocurrencies and blockchains has changed this. A blockchain is a distributed system for executing and recording transactions, which is maintained by many nodes without a central authority. All nodes collaboratively validate the information to be included in the blockchain through cryptography and distributed consensus. Blockchains offer resilience and security based on the collective trust placed in the nodes maintaining it.This presentation will revisit protocols for Byzantine consensus and explore older and newer protocols that power blockchains. The talk also introduces Hyperledger Fabric, a modular and extensible blockchain platform that is developed open-source under the Hyperledger Project and which was originally contributed by IBM. Fabric introduces a novel architecture for building resilient distributed systems that differs from the conventional paradigm, in order to accommodate flexible trust models, to cope with non-determinism, and to prevent resource exhaustion. There are currently several hundred prototypes, proofs-of-concept, and production systems of distributed ledger technology that use Fabric as a platform for distributing trust over the Internet.
12:00 – 14:00: Lunch
14:00 – 14:50: Session 9. Byzantine Consensus to Blockchains 2 ?
Speaker: Christian Cachin
14:50 – 15:05: Break
15:05 – 15:55: Session 10. Distributed Data Analytics
Speaker: Michael Franklin
Modern data-driven applications require quick answers to sophisticated questions over large amounts of data. In some cases, such as IoT or cloud scenarios, the data is inherently geographically distributed and hence, analytics processing must respect such distribution. In others, such as Big Data applications, scale-out architectures are used to deal with the sheer volume of data, requiring efficient mechanisms for performing complex analytics across 10’s to 1000’s of nodes. And in still other cases, answers are gathered and combined from disparate systems, such as when multiple machine learning systems are ensembled to produce more accurate predictions. When combined with requirements for low latency, high availability, and fault tolerance, the challenges in designing analytics systems for such environments are significant. In this lecture, we’ll discuss architectures and approaches for distributed analytics including parallel and distributed query processing, map-reduce architectures such as Spark, and streaming and continuous analytics systems.
15:55 – 18:00: Break
18:00 – 20:00: Poster Session (Styrhuset 1 & 2)
20:00 – 23:00: Banquet Dinner (Kaihuset)
Friday, August 17, 2018
04:30-05:30 Shuttle to Airport
Location: Outside the hotel main entrance.
For participants departing on flights SK4403, SK4411, and SK4419.
07:30 – 09:30: Breakfast and hotel checkout
09:30 – 10:20: Session 11. Security from Tags 1
Speaker: Fred Schneider
Reference monitors typically enforce security policies by intercepting operation invocations – the policy to be enforced is decomposed into operation-specific checks. These lectures discuss a data-centric alternative. Here, labels are attached to data, and each label gives a policy that describes how the associated value may be used. A new class of reactive information flow (RIF) labels is needed to fully support this view. We discuss the design of these RIF labels, give examples of their use, and describe a static enforcement scheme. We also discuss the need for label chains when run-time enforcement is employed, and we give some foundational results to characterize trade-offs. Joint work with Elisavet Kozyri.
10:20 – 10:40: Break and hotel checkout
10:40 – 11:30: Session 12. Security from Tags 2
Speaker: Fred Schneider
11:30 – 12:40: Lunch
12:40-13:40 Shuttle to Airport
Location: Outside the hotel main entrance.
For participants departing flights DY381 and SK4423.
12:40 – 13:30: Session 13. Reliable Distributed Storage 1
Speaker: Idit Keidar
With the increase in storage capacity demands, scalable storage solutions are increasingly adopting distributed storage solutions, where storage nodes communicate over a network. This talk will discuss the principles of building fault-tolerant distributed storage. We will discuss the ABD algorithm for emulating reliable shared storage using fail-prone storage nodes and the quorum-replication approach. We will continue to show how consensus (Disk Paxos) can be used for state-machine replication in such settings. Finally, we will touch on atomic transactions over shared storage.
13:30 – 13:40: Break
13:40 – 14:30: Session 14. Reliable Distributed Storage 2
Speaker: Idit Keidar