Title: Is TM still a research toy?
Name: Rachid Guerraoui

Abstract: This talk will address the question of whether there are still interesting research problems around transactional memory.

Title: JPaxos and Paxos STM
Name: Pawel Wojciechowski
Abstract: In the talk, I briefly describe two tools that we develop for implementing robust network services in Java. They include JPaxos -- an implementation of the State Machine Replication using the Lamport's MultiPaxos protocol and Paxos STM -- an object-oriented distributed transactional memory (which is based on JPaxos). In the talk I briefly describe the architecture of these systems and some evaluation results. Our research within this project spans a range of topics, including distributed algorithms, semantics, and static analysis.

Title: On the Cost of Concurrency in Transactional Memory
Name: Petr Kutznetsov
Abstract: The crux of software transactional memory (STM) is to combine an easy-to-use programming interface with an efficient utilization of the concurrent computing abilities provided by modern machines. But does this combination come with an inherent cost?
We evaluate the cost of concurrency by measuring the amount of expensive synchronization that must be employed in an STM implementation that ensures positive concurrency, i.e., allows for concurrent transaction processing in some executions. We focus on two popular progress conditions that provide positive concurrency: progressiveness and permissiveness.
We show that in permissive STMs, arguably providing the maximal degree of concurrency, the number of expensive synchronization patterns needed to implement a transaction is proportional to the size of the transaction's data set. In contrast, we show that progressive STMs that only guarantee minimal positive concurrency can be implemented using just a constant amount of synchronization. Then we show that even in progressive STMs, however, a transaction has to "protect" (e.g., by using locks or strong synchronization primitives) a linear amount of data with respect to its write-set size.

Title: Relaxed transactional access models
Name: Vasu Singh
Abstract: Transactional memory (TM) has shown potential to simplify the task of writing concurrent programs. Recent studies have shown that the conflict detection mechanisms for TM do not allow the performance of TM to scale to that of fine-grained locking in concurrent data structures. This has led researchers to relax the conflict detection for transactions, using mechanisms like early release and elastic transactions. We formalize the notion of relaxed transactional access models and their desired properties. We introduce a new transactional access model, peek-quake model,
that allows two new commands, peeks (a lighter form of read) --and its dual quakes, (a heavier form of write)-- for finer conflict detection in STM. In our augmented model, a normal read conflicts with every write, while a peek conflicts only with quakes. We compare the peek-quake access model to existing access models in our formalism. We use the peek-quake access model to implement concurrent data structures like linked list based sets and red-black trees.

Title: On Multi-Versioning in Transactional Memory
Name: Idit Keidar
Abstract: An effective way to reduce the number of aborts in STM is to keep multiple versions of transactional objects. We study inherent properties of STMs that use multiple versions to guarantee successful commits of all read-only transactions.
We show that some other desirable properties must be sacrificed if multi-versioning is used to ensure that all read only transactions commit successfully. In particular, we show that managing the versions (deciding which versions to keep and which can be garbage collected) is a major challenge.
We solve this challenge in a real system implementation -- we build a selective multi-versioning (SMV) algorithm. We demonstrate that an approach of keeping constant number of versions wastes too much memory (exponential memory growth in the worst case) and reaps too little benefits. In contrast, SMV keeps old object versions as long as they might be useful. Our algorithm succeeds to do so while keeping read-only transactions invisible by leveraging automatic garbage collection presented in Java. This way, SMV achieves much better (up to 11) throughput improvement over a single-version STM, while operating successfully in systems with limited memory.
Based on joint work with Dmitri Perelman and Rui Fan.

Title: On Using Separation Logic to Detect Snapshot Isolation Anomalies in Software Transactional Memory
Name: Ricardo Dias
Abstract: Some performance issues of software transactional memory are caused by unnecessary abort situations where non serializable and yet non conflicting transactions are scheduled to execute concurrently. By smartly relaxing the isolation properties of transactions, transactional systems may overcome these performance issues and attain considerable gains. However, it is known that relaxing isolation restrictions may lead to execution anomalies causing programs to yield unexpected results. Some database management systems make that compromise and allow the option to operate with relaxed isolation mechanisms. In these cases, developers must accept the fact that they must explicitly deal with anomalies. In software transactional memory systems, where transactions protect the volatile state of programs, execution anomalies may have severe and unforeseen semantic consequences. In this setting, the balance between a relaxed isolation level and the ability to enforce the necessary correctness in general purpose language programs is harder to achieve.
The solution we devise in this work is to statically analyze programs, using Separation Logic, to detect the kind of execution anomalies that emerge under snapshot isolation. Our approach allows a compiler to either warn the developer about the possible snapshot isolation anomalies in a given program, or to possibly inform automatic correctness strategies to ensure execution under full serializability.

Title: Bounds on Transactional Scheduling
Name: Alessia Milani
Abstract: The transactional approach to contention management guarantees atomicity by aborting transactions that may violate consistency. A major challenge in this approach is to schedule transactions in a manner that reduces the total time to perform all transactions (the makespan), since transactions are often aborted and restarted. The performance of a transactional scheduler can be evaluated by the ratio between its makespan and the makespan of an optimal, clairvoyant scheduler that knows the list of resource accesses that will be performed by each transaction, as well as its release time and duration.
I will present some bounds on transactional scheduling for read-dominated workloads which include read-only transactions and transactions that write only towards the end of their execution.  Finally, I will present the BIMODAL transactional scheduler, which is especially tailored to accommodate read-only transactions, without punishing transactions that write most of their duration.

Title: On Proving Isolation Properties for Software Transactional Memory
Name: Annette Bieniusa
Abstract: An algorithm for STM is correct if it guarantees a proclaimed degree of isolation between concurrently executing transactions. A correctness proof requires explicit modeling of the effects of transaction bodies and the non-deterministic scheduling of their operations. We provide a formalization of STM that is explicit about all aspects required for a correctness proof: effects of operations, non-determinism, and modeling rollback. We prove that this algorithm is correct by showing that it implements opacity.

Title: Towards a universal construction for transaction-based multiprocess programs
Name: Michel Raynal
Abstract: The aim of a Software Transactional Memory (STM) system is to discharge the programmer from the explicit management of synchronization issues. The programmer's job resides in the design of multiprocess programs in which processes are made up of transactions, each transaction being an atomic execution unit that accesses concurrent objects. The important point is that the programmer has to focus her/his efforts only on the parts of code which have to be atomic execution units without worrying on the way the corresponding synchronization has to be realized. Non-trivial STM systems allow transactions to execute concurrently and rely on the notion of commit/abort of a transaction in order to solve their conflicts on the objects they access simultaneously. In some cases, the management of aborted transactions is left to the programmer. In other cases, the underlying system scheduler is appropriately modified or an underlying contention manager is used in order that each transaction be (``practically always'' or with high probability) eventually committed. This talk will present a deterministic STM system in which (1) every invocation of a transaction is executed exactly once and (2) the notion of commit/abort of a transaction remains unknown to the programmer. This system, which imposes restriction neither on the design of processes nor or their concurrency pattern, can be seen as a step in the design of a deterministic universal construction to execute transaction-based multiprocess programs on top of a multiprocessor. Interestingly, the proposed construction is lock-free (in the sense that it uses no lock).

Talks for WG3

Title: TMBox: A Configurable 16-core Hybrid TM FPGA prototype
Name: Osman Unsal
Abstract: In this talk, I will introduce TMBox, our open-source 16-core FPGA hybrid prototype. TMBox is built from a publicly available MIPS R2000 soft-core. We heavily modified this core to make it suitable to function in a multi-processor environment. These modifications include TLB and MMU support, floating-point units, exception support, synchronization primitives: LL/SC, snoopy caches, MSI cache coherence, performance counters and system libraries to support string and I/O operations. We then implemented a per-core hardware TM unit by extending the cache finite state machine, a tx-cache and new TM instruction extensions. Our design closely follows the TinySTM-ASF proposal: I will mainly talk about the choices we took while designing TMBox and conclude with our initial set of results.

Title: The Velox Transactional Memory Stack
Name: Patrick Marlier
Abstract: The adoption of multi- and many-core architectures for mainstream computing undoubtedly brings profound changes in the way software is developed. In particular, the use of fine grained locking as the multi-core programmer's coordination methodology is considered by more and more experts as a dead-end. The transactional memory (TM) programming paradigm is a strong contender to become the approach of choice for replacing locks and implementing atomic operations in concurrent programming. Combining sequences of concurrent operations into atomic transactions allows a great reduction in the complexity of both programming and verification, by making parts of the code appear to execute sequentially without the need to program using fine-grained locking. Transactions remove from the programmer the burden of figuring out the interaction among concurrent operations that happen to conflict when accessing the same locations in memory. The EU-funded FP7 VELOX project designs, implements and evaluates an integrated TM stack, spanning from programming language to the hardware support, and including runtime and libraries, compilers, and application environments. This presentation will present an overview of the VELOX TM stack and its associated challenges and contributions.

Talks for WG4

Title: MUTS: Native Scala Constructs for Software Transactional Memory
Name: Daniel Goodman
Abstract: In this paper we argue that the current approaches to imple- menting transactional memory in Scala, while very clean, adversely affect the programmability, readability and main- tainability of transactional code. These problems occur out of a desire to avoid making modifications to the Scala com- piler. As an alternative we introduce Manchester University Transactions for Scala (MUTS), which instead adds key- words to the Scala compiler to allow for the implementation of transactions through traditional block syntax such as that used in “while” statements. This allows for transactions that do not require a change of syntax style and do not restrict their granularity to whole classes or methods. While imple- menting MUTS does require some changes to the compiler’s parser, no further changes are required to the compiler. This is achieved by the parser describing the transactions in terms of existing constructs of the abstract syntax tree, and the use of Java Agents to rewrite to resulting class files once the compiler has completed.

Title: Atomicity and Structured Concurrency
Name: David Kitchin
Abstract: Concurrent programming is notoriously difficult. By its nature, concurrency introduces an entirely new set of possible program behaviors about which the programmer must reason. Current programming abstractions have not been adequate to manage this complexity. Threads with shared state are at present the dominant approach to concurrent programming. Under this model, the dynamic structure of the concurrent execution is not at all apparent from the program's lexical structure. Message passing models alleviate many of these difficulties, but the global model is still unstructured. Since the program is factored into individual actors and their interactions, it remains difficult to see the overall structure of the execution.
The Orc programming language changes the conceptual model of programming to be inherently concurrent in a structured way. The concurrent structure of an Orc execution is evident from the lexical structure of the program, improving on both the shared state model and the message passing model. Orc uses sites to communicate with the external world. Sites are numerous and diverse, they can have nontrivial semantics, and in principle they each represent some kind of shared resource. Thus, the difficulties inherent in using shared state still recur frequently even when writing small Orc programs.
At present, Orc programs make use of locks, semaphores, and other mutual exclusion disciplines to safeguard access to shared state at sites. This gives rise to all of the difficulties inherent in locking. Additionally, Orc sites may represent external services, which may be concurrently interacting with other clients that might not be aware of the locking discipline. Thus, program-level locks are insufficient to ensure safety.
The use of language-level transactions, especially in the context of transactional memory, has been explored as an alternative to locks. Inspired by this approach, I propose Orca, an extension of Orc with a new 'atomic' combinator. 'atomic' executes an expression in such a way that: (1) from the expression's perspective, no concurrent events take place outside of it during its execution, and (2) from any external execution's perspective, the expression evaluates in its entirety in one atomic step. I will show how this notion of atomicity can be formalized simply, symmetrically, and without appeal to linearization or traces.
The 'atomic' combinator has two critical characteristics. First, it ensures atomicity for all events, not just memory accesses; every site called within 'atomic' is invoked in an atomic context, and each such site must determine internally how to maintain the atomicity of each such call with respect to its other clients, including its other external clients. Second, 'atomic' can be recursively nested under any of the combinators, such as '|' (parallel evaluation), '<<' (fork-join), or even 'atomic' itself. This allows unlimited concurrent evaluation inside of transactions, as well as unbounded nesting of such parallel transactions.
I will also briefly discuss the comparative expressivity of Orca, and the pi-calculus with guarded choice, introducing a pattern called "atomic choice" using the atomic combinator.

Title: Integration of transactional memory support into a data-flow extension of OpenMP
Name: Ismail Kuru
Abstract: The combination of data-flow programming with transactional memory is one ambitious challenge addressed by the TERAFLUX European project, aiming increased expressiveness and performance while preserving the paradigms' properties. To explore different semantics, compilation and runtime methods, we extend the OpenMP specification and implement prototypes based on GCC. With the maturation of these prototypes, time has come to experiment it together with the emerging transactional memory branch of GCC. This project combines development and research components  and can be decomposed into 5 phases.
1. Study the compatibility of the transactional memory and OpenMP constructs in the transmem development branch of GCC and propose solutions to the possible technical difficulties.
2. Study the design and implementation of both data-flow (streaming) and transmem branches of GCC, interacting with their maintainers.
3. Address the infrastructure limitations, software engineering and version management issues related to the integration of both code bases into a single experimental branch.
4. Contribute to the design and implementation for a semantics of transactions nested within OpenMP tasks, currently being defined in the TERAFLUX project.
5. Write and rewrite relevant benchmarks leveraging the new programming model, performing systematic evaluations and detailed characterizations of the performance behavior.The first two phases would be assigned as a research project during Ismail Kuru's curriculum at T.U. München. While the last three phases would constitute the matter of his Master thesis work during the Summer, working at INRIA in Paris.

Title: Programmer and Runtime Policies for Adaptive Many-Core Parallelism
Name: Jeremy Singer
Abstract: This project investigates a novel extension to the Java programming language, providing support for flexible parallelism on next-generation many-core processors. The flexibility of our system is based on: (1) encoding descriptions of the level of accuracy provided by alternative implementations of a program module, allowing trade-offs between computational cost and precision; and (2) enabling the runtime system to determine adaptively when new parallel threads should be spawned. I argue that this kind of programming extension fits naturally in a transactional memory system setting.

Title: Methods and Strategies to Rate and Optimize Transactional Memory Applications
Name: Martin Schindewolf
Abstract: This talk will give insights on the goals and ongoing work in the DFG-funded project ``TM-Opt''. The main goal of this project is to develop strategies that enable the rating and optimization of transactional memory applications. The research in this project aims to complement and enrich ongoing transactional memory research by adopting the view of the programmer. Through tool support, the project aims to close the gap between low-level TM research, such as TM models and conflict detection algorithms, and the high-level view of the programmer.
The strategies enable to rate and optimize the run time behavior of TM applications. As a first step, the run time behavior needs to be captured and preserved. Second, a post-processing step enables to analyze and rate the run time behavior. Through a customized analysis, bottlenecks and repeatedly conflicting transactions are identified. The TM tools highlight such performance-critical situations and guide the programmer to solve these.

Title: Crafting a High-Performance Ready-to-Go STM
Name: Guy Korlan
Abstract: The goal of this talk is to show that an STM having all the necessary real-world features with can be built without compromising performance. For that we will present the following directions: A Java STM based on state-of-the-art algorithms. Built as an open research platform for new algorithms and methods. Static analysis optimizations to the platform. Hand craft most used libraries to be STM Ready. Benchmarks STM with real world applications. The basis of our research initiative is Deuce, our novel open-source Java framework for transactional memory. Deuce has several desired features not found in earlier Java STM frameworks. Currently there does not exist an efficient Java STM framework that delivers a full featured STM that can be added to an existing application without changes to its compiler or libraries. It was not clear if one could build an efficient Java STM without compiler support. Deuce is intended to fill this gap. It is non-intrusive in the sense that no modifications to the Java virtual machine (JVM) or extensions to the language are necessary.

Title: JVSTM and its applications
Name: Joao Cachopo
Abstract: The purpose of my presentation is to give an overview of the work that we have been doing since 2004 in the development of JVSTM, a Java library that implements a multi-version STM, and our experience in applying it to various domains. JVSTM was extended to support persistence and is being used in a real-world large web application since 2005, being the first known use of an STM in a production environment. Moreover, we have applied it to several domains, from automatic memoization, to dynamic software upgrades, passing by automatic parallelization of programs through thread-level speculation, which I will discuss briefly.

Talks for WG5

Title: CloudTM project: a distributed transactional memory platform for the Cloud
Name: Paolo Romano
Abstract: Cloud Computing has emerged as a new paradigm for deploying, managing and offering services through a shared infrastructure. The foreseen benefits of Cloud Computing are very compelling both from a cloud consumer and from a cloud services provider perspective: freeing corporations from large IT capital investments via usage-based pricing schemes; leveraging the economies of scale for both services providers and users of the cloud; easing the deployment of services.
One of the main challenges to materialize these perceived benefits is to identify innovative distributed programming models that simplify the development of Cloud-based services, allowing for ordinary programmers to take full advantage of the seemingly unbounded amount of computational power and storage available on demand in large scale Cloud infrastructures.
This project aims at designing, building, and evaluating an innovative middleware platform for service implementation of Cloud-based services: Cloud-TM (Cloud-Transactional Memory). Cloud-TM offers a simple and intuitive programming model for large scale distributed applications that integrates the familiar notion of atomic transaction as a first-class programming language construct, sparing programmers from the burden of implementing low level, error-prone mechanisms (e.g. locking, persistence and fault-tolerance) and permitting major reductions in the time and cost of the development process.
Cloud-TM will embed a set of autonomic mechanisms to simplify service monitoring and administration, a major source of costs in dynamic and elastic environments such as the cloud. These mechanisms aim at ensuring the achievement of user defined Quality of Service levels at minimum operational costs by automating the provisioning of resources from the cloud and self-tuning the middleware platform to achieve optimal efficiency in the utilization of resources.

Title: Transactional Resources in Cloud Environment
Name: Antonio Puliafito
Abstract: Resources provisioning is a very important issue in cloud computing, i.e. how resources may be organized, allocated and managed in order to deliver high quality services. Clouds make wide use of virtual resources, which are usually managed in terms of Virtual Machines (VMs) and/or Virtual Clusters (VCs). In a cloud environment, the entity that performs resource provisioning, called Resource Manager (RM), has to allocate resources for applications/services on VMs or VCs, which may be deployed on several physical servers in the same cloud, or even in different clouds. Then, a RM has to maintain access to such resources until the applications/services are terminated, executing several tasks such as:
- collecting and indexing all the resources available from cloud providers;
- evaluating the most suitable VMs or VCs for the application needs;
- allocating specific resources to each user/application request in order to ensure agreed requirements of QoS ;
- monitoring the usage of allocated resources in order to verify the SLA with the user;
- dynamicalmly reallocating resources, so to have a more efficiently use of available resources;
- when Application/Services request to allocate (extra) resources, checking whether they are authorized;
- managing VMs migration among different physical sites.
We believe that Transactional Memories represent a very helpful technology to implement all the functionalities necessary in a cloud architecture to manage virtual resources. In particular, our interest is focused on Software TM to support the implementation of the RM entity.
Transactions can provide a lightweight mechanism to synchronize all the activities on the virtual resources, alleviating many of the problems associated with the locking of resources. In particular, the RM can very efficiently deal with different requests from users or applications, since the resource allocation tasks are executed atomically. They are able to guarantee the consistency of the shared virtual resources, thereby reducing the complexity of both programming and verification.
Using transactions, there is no need to lock the virtual resources during the decision tasks, before allocating them or to unlock if the user request can not be satisfied. In fact, the RM could spend a lot of time before deciding how to settle a specific request, due to several factors, such as discovery of resources able to fulfill the request, priority in request queues, renegotiation of SLA.
Furthermore, in classic lock-based parallel programming, the order of the locks is very important to avoid deadlocks and livelocks. Hence transactional schemes can be used to develop more robust and flexible code.
Transactions can help to improve the efficiency of functionalities inside the RM, but also to synchronize different entities that affect the composition of virtual resources. The availability of cloud providers and their related resources can change during the time, according to agreements between the cloud manager and cloud providers themselves. This is especially true in cloud environments based on volunteered resources. In fact, volunteered cloud systems use computers voluntarily offered by their owners. So, user resources are not only a passive interface to cloud-based services, but can actively participate with other Clouds to create a larger infrastructure. However, in such a type of scenario, the availability of resources can not be guaranteed during the time and the RM has to interact with a very dynamic virtual infrastructure. For example, if in the meantime the RM is allocating a specific resource and the related cloud provider modifies its availability in resource provisioning or the related volunteered provider abruptly leaves the system, the allocation process in the RM will be affected by inconsistency problems. However, by using transactions, the allocation process in the RM will abort and start again, solving the inconsistency problem.
As additional advantage, transaction priorities offer a very natural way to manage requests characterized by different SLA and with different priority levesl. In fact, it serves as a contention management strategy that allows the RM to determine which of two transactions should abort if they encounter a conflict on a virtual resource. Lower priority transactions should defer to higher priority ones.
The application of transactional memories in cloud environment is not trivial and several challenging issues need to be investigated. First, an appropriate representation of virtual resources is necessary, since the introduction of transactions to an application can have negative, counterintuitive consequences for performance if data structures are badly designed. Second, due to the high number and complexity of functionalities that operate on virtual resources, it is very hard to specify how to divide them into atomic transactions. Specific atomization strategies should be developed, because threads with very long execution time cause a lot of redundancy and overhead if they incur in aborting.
With regard transactions aborts, we know that detecting conflicts increases the transaction throughput, as transactions do not perform useless work. Commit-time locking may help to avoid some read-write conflicts, but in general conflicts discovered at commit time cannot be solved without aborting at least one transaction. We guess that in cloud environment conflicts on virtual resources are less probable, since a successful deployment of a Cloud environment implies the collection of a huge amount of virtual resources. However, due the high dynamism of the cloud environment, we have to plan a strategy to overcome this potential situation. Thus, a run-time conflict density evaluation has to be performed, in order to asses how many events are likely to be involved in a virtual resource conflict. Then, specific policies should be implemented to limit the number of concurrently executing transactions at a given time if the conflict density is very high.
We are investigating all these issues with reference to a specific cloud architecture, that is Cloud@Home (https://cloudathome.unime.it). It is a new volunteer computing paradigm, where the infrastructure is obtained by merging heterogeneous resources offered by different domains and/or providers such as other Clouds, Grid farms, clusters, and datacenters, till single desktops.

Title: Dynamic replicas of parallel stream applications in interconnection networks
Name: Dimiter Avresky
Abstract: The recent architectural trend that has lead to the widespread adoption of multicore CPUs has fostered a huge research interest to wards Software Transactional Memory (STM). As STMs are starting to face the high availability and scalability requirements of real world production environments, it is natural to foresee the need for repli cation solutions specifically tailored for STMs. It has been demonstrated that Motion JPEG2000 application can be parallelized and formally represented as a Directed Acyclic Graph (DAG) and adaptively mapped on the interconnection network. It is worth noting that the proposed technique has been validated by simulation of 1000 cores 2D mesh topology for random patterns in the presence of failures up to 10%. The hierarchical model is utilized, based on DAG (fork - join) graph presentation of parallel streaming application (Motion JPEG), for keeping dynamically the replicas of the children in the Stream Leader(s). This model is reducing drastically the latency for spreading replicas in distributed systems (on memory or hard drives.) Based on a learning technique, the "explicit path" is created for each stream (atomic transaction). The method tolerates multiple failures (stream leaders and children ) in DAG graph. In network cache coherence, by utilizing Virtual Tree Cache, will be exploited for reducing the communication latency.

Title: Performance Modelling and Replication of Software Transactional Memories
Name: Pierangelo Di Sanzo and Roberto Palmieri
Abstract: Software Transactional Memories (STMs) are emerging as a potentially disruptive programming model. Within this context, the activities of our research group is focused on two aspects. The first one deals with dependability of STM systems via replication, an the main target of our work has been to maximize the overlap between communication/coordination and local processing, essentially by means of speculative processing techniques. The key idea behind the proposed approaches is to explore the optimistic delivery event offered by the Optimistic Atomic Broadcast (OAB) group communication protocol, to speculatively process transactions in tentative orders, and propagate uncommitted data versions along conflicting transactions chains. The second one deals with performance evaluation of STM systems. As for this aspect, these systems are based on (and require) innovative design and development approaches, where the optimization focus is shifted on aspects that historically had less importance in  traditional transactional systems (e.g. DBMS). Following these  considerations, the our main research target has been to cope with such  peculiarities, providing performance evaluation tools that well fit them.

Title: Distributed Transactional Memory for Clusters and Grids
Name: Michael Schoettner
Abstract:This talk will present distributed transactional memory (DTM) experiences from two projects and gives an outlook on perspectives for DTM in cloud computing applications. The Plurix project (1996-2004) implements a DTM for PC clusters. The central abstraction of this language-based operating system (OS) is a DTM storing data and code (even the OS code and drivers are stored in the DTM). Applications are based on an event-driven architecture where events are handled using speculative transactions. Furthermore, we have implemented another DTM within the Object Sharing Service (OSS) part of the XtreemOS project funded by the EU (FP6, 2006-2010). OSS implements a shared library for the Linux-based XtreemOS grid system with transparent data access detection, adaptive conflict unit size management, and various commit protocols. The algorithms use explicitly marked transaction boundaries and have been evaluated using a MMVE, a ray tracer and micro benchmarks. Currently, we are adopting DTM concepts for an extended MapReduce framework and MMVEs running in cloud environments.