arXiv subjects

Gene Cooperman

Publications and source records attributed to Gene Cooperman.

36 records · Page 2Linked to original sources

arXiv2017

Transition Watchpoints: Teaching Old Debuggers New Tricks

Reversible debuggers and process replay have been developed at least since 1970. This vision enables one to execute backwards in time under a debugger. Two important problems in practice are that, first, current reversible debuggers are slow when reversing over long time periods, and, second, after building one reversible debugger, it is difficult to transfer that achievement to a new programming environment. The user observes a bug when arriving at an error. Searching backwards for the corresponding fault may require many reverse steps. Ultimately, the user prefers to write an expression that will transition to false upon arriving at the fault. The solution is an expression-transition watchpoint facility based on top of snapshots and record/replay. Expression-transition watch- points are implemented as binary search through the timeline of a program execution, while using the snapshots as landmarks within that timeline. This allows for debugging of subtle bugs that appear only after minutes or more of program execution. When a bug occurs within seconds of program startup, repeated debugging sessions suffice. Reversible debugging is preferred for bugs seen only after minutes. This architecture allows for an efficient and easy-to-write snapshot-based reversibe debugger on top of a conventional debugger. The validity of this approach was tested by developing four personalities (for GDB, MATLAB, Perl, and Python), with each personality typically requiring just 100 lines of code.

cs.PL↗

arXiv2017

Adapting the DMTCP Plugin Model for Checkpointing of Hardware Emulation

Checkpoint-restart is now a mature technology. It allows a user to save and later restore the state of a running process. The new plugin model for the upcoming version 3.0 of DMTCP (Distributed MultiThreaded Checkpointing) is described here. This plugin model allows a target application to disconnect from the hardware emulator at checkpoint time and then re-connect to a possibly different hardware emulator at the time of restart. The DMTCP plugin model is important in allowing three distinct parties to seamlessly inter-operate. The three parties are: the EDA designer, who is concerned with formal verification of a circuit design; the DMTCP developers, who are concerned with providing transparent checkpointing during the circuit emulation; and the hardware emulator vendor, who provides a plugin library that responds to checkpoint, restart, and other events. The new plugin model is an example of process-level virtualization: virtualization of external abstractions from within a process. This capability is motivated by scenarios for testing circuit models with the help of a hardware emulator. The plugin model enables a three-way collaboration: allowing a circuit designer and emulator vendor to each contribute separate proprietary plugins while sharing an open source software framework from the DMTCP developers. This provides a more flexible platform, where different fault injection models based on plugins can be designed within the DMTCP checkpointing framework. After initialization, one restarts from a checkpointed state under the control of the desired plugin. This restart saves the time spent in simulating the initialization phase, while enabling fault injection exactly at the region of interest. Upon restart, one can inject faults or otherwise modify the remainder of the simulation. The work concludes with a brief survey of checkpointing and process-level virtualization.

cs.OS↗

arXiv2016

System-level Scalable Checkpoint-Restart for Petascale Computing

Fault tolerance for the upcoming exascale generation has long been an area of active research. One of the components of a fault tolerance strategy is checkpointing. Petascale-level checkpointing is demonstrated through a new mechanism for virtualization of the InfiniBand UD (unreliable datagram) mode, and for updating the remote address on each UD-based send, due to lack of a fixed peer. Note that InfiniBand UD is required to support modern MPI implementations. An extrapolation from the current results to future SSD-based storage systems provides evidence that the current approach will remain practical in the exascale generation. This transparent checkpointing approach is evaluated using a framework of the DMTCP checkpointing package. Results are shown for HPCG (linear algebra), NAMD (molecular dynamics), and the NAS NPB benchmarks. In tests up to 32,752 MPI processes on 32,752 CPU cores, checkpointing of a computation with a 38 TB memory footprint in 11 minutes is demonstrated. Runtime overhead is reduced to less than 1%. The approach is also evaluated across three widely used MPI implementations.

cs.DC↗

arXiv2015

Checkpointing as a Service in Heterogeneous Cloud Environments

A non-invasive, cloud-agnostic approach is demonstrated for extending existing cloud platforms to include checkpoint-restart capability. Most cloud platforms currently rely on each application to provide its own fault tolerance. A uniform mechanism within the cloud itself serves two purposes: (a) direct support for long-running jobs, which would otherwise require a custom fault-tolerant mechanism for each application; and (b) the administrative capability to manage an over-subscribed cloud by temporarily swapping out jobs when higher priority jobs arrive. An advantage of this uniform approach is that it also supports parallel and distributed computations, over both TCP and InfiniBand, thus allowing traditional HPC applications to take advantage of an existing cloud infrastructure. Additionally, an integrated health-monitoring mechanism detects when long-running jobs either fail or incur exceptionally low performance, perhaps due to resource starvation, and proactively suspends the job. The cloud-agnostic feature is demonstrated by applying the implementation to two very different cloud platforms: Snooze and OpenStack. The use of a cloud-agnostic architecture also enables, for the first time, migration of applications from one cloud platform to another.

cs.DC↗

arXiv2014

Transparent Checkpoint-Restart for Hardware-Accelerated 3D Graphics

Providing fault-tolerance for long-running GPU-intensive jobs requires application-specific solutions, and often involves saving the state of complex data structures spread among many graphics libraries. This work describes a mechanism for transparent GPU-independent checkpoint-restart of 3D graphics. The approach is based on a record-prune-replay paradigm: all OpenGL calls relevant to the graphics driver state are recorded; calls not relevant to the internal driver state as of the last graphics frame prior to checkpoint are discarded; and the remaining calls are replayed on restart. A previous approach for OpenGL 1.5, based on a shadow device driver, required more than 78,000 lines of OpenGL-specific code. In contrast, the new approach, based on record-prune-replay, is used to implement the same case in just 4,500 lines of code. The speed of this approach varies between 80 per cent and nearly 100 per cent of the speed of the native hardware acceleration for OpenGL 1.5, as measured when running the ioquake3 game under Linux. This approach has also been extended to demonstrate checkpointing of OpenGL 3.0 for the first time, with a demonstration for PyMol, for molecular visualization.

cs.OS↗

arXiv2014

Transparent Checkpoint-Restart over InfiniBand

InfiniBand is widely used for low-latency, high-throughput cluster computing. Saving the state of the InfiniBand network as part of distributed checkpointing has been a long-standing challenge for researchers. Because of a lack of a solution, typical MPI implementations have included custom checkpoint-restart services that "tear down" the network, checkpoint each node as if the node were a standalone computer, and then re-connect the network again. We present the first example of transparent, system-initiated checkpoint-restart that directly supports InfiniBand. The new approach is independent of any particular Linux kernel, thus simplifying the current practice of using a kernel-based module, such as BLCR. This direct approach results in checkpoints that are found to be faster than with the use of a checkpoint-restart service. The generality of this approach is shown not only by checkpointing an MPI computation, but also a native UPC computation (Berkeley Unified Parallel C), which does not use MPI. Scalability is shown by checkpointing 2,048 MPI processes across 128 nodes (with 16 cores per node). In addition, a cost-effective debugging approach is also enabled, in which a checkpoint image from an InfiniBand-based production cluster is copied to a local Ethernet-based cluster, where it can be restarted and an interactive debugger can be attached to it. This work is based on a plugin that extends the DMTCP (Distributed MultiThreaded CheckPointing) checkpoint-restart package.

cs.OS↗

arXiv2014

Use of checkpoint-restart for complex HEP software on traditional architectures and Intel MIC

Process checkpoint-restart is a technology with great potential for use in HEP workflows. Use cases include debugging, reducing the startup time of applications both in offline batch jobs and the High Level Trigger, permitting job preemption in environments where spare CPU cycles are being used opportunistically and efficient scheduling of a mix of multicore and single-threaded jobs. We report on tests of checkpoint-restart technology using CMS software, Geant4-MT (multi-threaded Geant4), and the DMTCP (Distributed Multithreaded Checkpointing) package. We analyze both single- and multi-threaded applications and test on both standard Intel x86 architectures and on Intel MIC. The tests with multi-threaded applications on Intel MIC are used to consider scalability and performance. These are considered an indicator of what the future may hold for many-core computing.

physics.comp-ph↗

arXiv2014

Explorations of the viability of ARM and Xeon Phi for physics processing

We report on our investigations into the viability of the ARM processor and the Intel Xeon Phi co-processor for scientific computing. We describe our experience porting software to these processors and running benchmarks using real physics applications to explore the potential of these processors for production physics processing.

physics.comp-ph↗

arXiv2012

FReD: Automated Debugging via Binary Search through a Process Lifetime

Reversible debuggers have been developed at least since 1970. Such a feature is useful when the cause of a bug is close in time to the bug manifestation. When the cause is far back in time, one resorts to setting appropriate breakpoints in the debugger and beginning a new debugging session. For these cases when the cause of a bug is far in time from its manifestation, bug diagnosis requires a series of debugging sessions with which to narrow down the cause of the bug. For such "difficult" bugs, this work presents an automated tool to search through the process lifetime and locate the cause. As an example, the bug could be related to a program invariant failing. A binary search through the process lifetime suffices, since the invariant expression is true at the beginning of the program execution, and false when the bug is encountered. An algorithm for such a binary search is presented within the FReD (Fast Reversible Debugger) software. It is based on the ability to checkpoint, restart and deterministically replay the multiple processes of a debugging session. It is based on GDB (a debugger), DMTCP (for checkpoint-restart), and a custom deterministic record-replay plugin for DMTCP. FReD supports complex, real-world multithreaded programs, such as MySQL and Firefox. Further, the binary search is robust. It operates on multi-threaded programs, and takes advantage of multi-core architectures during replay.

cs.SE↗

arXiv2012

A Generic Checkpoint-Restart Mechanism for Virtual Machines

It is common today to deploy complex software inside a virtual machine (VM). Snapshots provide rapid deployment, migration between hosts, dependability (fault tolerance), and security (insulating a guest VM from the host). Yet, for each virtual machine, the code for snapshots is laboriously developed on a per-VM basis. This work demonstrates a generic checkpoint-restart mechanism for virtual machines. The mechanism is based on a plugin on top of an unmodified user-space checkpoint-restart package, DMTCP. Checkpoint-restart is demonstrated for three virtual machines: Lguest, user-space QEMU, and KVM/QEMU. The plugins for Lguest and KVM/QEMU require just 200 lines of code. The Lguest kernel driver API is augmented by 40 lines of code. DMTCP checkpoints user-space QEMU without any new code. KVM/QEMU, user-space QEMU, and DMTCP need no modification. The design benefits from other DMTCP features and plugins. Experiments demonstrate checkpoint and restart in 0.2 seconds using forked checkpointing, mmap-based fast-restart, and incremental Btrfs-based snapshots.

cs.OS↗

arXiv2011

A Bit-Compatible Shared Memory Parallelization for ILU(k) Preconditioning and a Bit-Compatible Generalization to Distributed Memory

ILU(k) is a commonly used preconditioner for iterative linear solvers for sparse, non-symmetric systems. It is often preferred for the sake of its stability. We present TPILU(k), the first efficiently parallelized ILU(k) preconditioner that maintains this important stability property. Even better, TPILU(k) preconditioning produces an answer that is bit-compatible with the sequential ILU(k) preconditioning. In terms of performance, the TPILU(k) preconditioning is shown to run faster whenever more cores are made available to it --- while continuing to be as stable as sequential ILU(k). This is in contrast to some competing methods that may become unstable if the degree of thread parallelism is raised too far. Where Block Jacobi ILU(k) fails in an application, it can be replaced by TPILU(k) in order to maintain good performance, while also achieving full stability. As a further optimization, TPILU(k) offers an optional level-based incomplete inverse method as a fast approximation for the original ILU(k) preconditioned matrix. Although this enhancement is not bit-compatible with classical ILU(k), it is bit-compatible with the output from the single-threaded version of the same algorithm. In experiments on a 16-core computer, the enhanced TPILU(k)-based iterative linear solver performed up to 9 times faster. As we approach an era of many-core computing, the ability to efficiently take advantage of many cores will become ever more important. TPILU(k) also demonstrates good performance on cluster or Grid. For example, the new algorithm achieves 50 times speedup with 80 nodes for general sparse matrices of dimension 160,000 that are diagonally dominant.

cs.DC↗

arXiv2011

Finding the Minimal DFA of Very Large Finite State Automata with an Application to Token Passing Networks

Finite state automata (FSA) are ubiquitous in computer science. Two of the most important algorithms for FSA processing are the conversion of a non-deterministic finite automaton (NFA) to a deterministic finite automaton (DFA), and then the production of the unique minimal DFA for the original NFA. We exhibit a parallel disk-based algorithm that uses a cluster of 29 commodity computers to produce an intermediate DFA with almost two billion states and then continues by producing the corresponding unique minimal DFA with less than 800,000 states. The largest previous such computation in the literature was carried out on a 512-processor CM-5 supercomputer in 1996. That computation produced an intermediate DFA with 525,000 states and an unreported number of states for the corresponding minimal DFA. The work is used to provide strong experimental evidence satisfying a conjecture on a series of token passing networks. The conjecture concerns stack sortable permutations for a finite stack and a 3-buffer. The origins of this problem lie in the work on restricted permutations begun by Knuth and Tarjan in the late 1960s. The parallel disk-based computation is also compared with both a single-threaded and multi-threaded RAM-based implementation using a 16-core 128 GB large shared memory computer.

cs.DC↗

arXiv2009

Temporal Debugging using URDB

A new style of temporal debugging is proposed. The new URDB debugger can employ such techniques as temporal search for finding an underlying fault that is causing a bug. This improves on the standard iterative debugging style, which iteratively re-executes a program under debugger control in the search for the underlying fault. URDB acts as a meta-debugger, with current support for four widely used debuggers: gdb, MATLAB, python, and perl. Support for a new debugger can be added in a few hours. Among its points of novelty are: (i) the first reversible debuggers for MATLAB, python, and perl; (ii) support for today's multi-core architectures; (iii) reversible debugging of multi-process and distributed computations; and (iv) temporal search on changes in program expressions. URDB gains its reversibility and temporal abilities through the fast checkpoint-restart capability of DMTCP (Distributed MultiThreaded CheckPointing). The recently enhanced DMTCP also adds ptrace support, enabling one to freeze, migrate, and replicate debugging sessions.

cs.OS↗

arXiv2009

DMTCP: Transparent Checkpointing for Cluster Computations and the Desktop

DMTCP (Distributed MultiThreaded CheckPointing) is a transparent user-level checkpointing package for distributed applications. Checkpointing and restart is demonstrated for a wide range of over 20 well known applications, including MATLAB, Python, TightVNC, MPICH2, OpenMPI, and runCMS. RunCMS runs as a 680 MB image in memory that includes 540 dynamic libraries, and is used for the CMS experiment of the Large Hadron Collider at CERN. DMTCP transparently checkpoints general cluster computations consisting of many nodes, processes, and threads; as well as typical desktop applications. On 128 distributed cores (32 nodes), checkpoint and restart times are typically 2 seconds, with negligible run-time overhead. Typical checkpoint times are reduced to 0.2 seconds when using forked checkpointing. Experimental results show that checkpoint time remains nearly constant as the number of nodes increases on a medium-size cluster. DMTCP automatically accounts for fork, exec, ssh, mutexes/semaphores, TCP/IP sockets, UNIX domain sockets, pipes, ptys (pseudo-terminals), terminal modes, ownership of controlling terminals, signal handlers, open file descriptors, shared open file descriptors, I/O (including the readline library), shared memory (via mmap), parent-child process relationships, pid virtualization, and other operating system artifacts. By emphasizing an unprivileged, user-space approach, compatibility is maintained across Linux kernels from 2.6.9 through the current 2.6.28. Since DMTCP is unprivileged and does not require special kernel modules or kernel patches, DMTCP can be incorporated and distributed as a checkpoint-restart module within some larger package.

cs.DC↗

arXiv2005

Fast Query Processing by Distributing an Index over CPU Caches

Data intensive applications on clusters often require requests quickly be sent to the node managing the desired data. In many applications, one must look through a sorted tree structure to determine the responsible node for accessing or storing the data. Examples include object tracking in sensor networks, packet routing over the internet, request processing in publish-subscribe middleware, and query processing in database systems. When the tree structure is larger than the CPU cache, the standard implementation potentially incurs many cache misses for each lookup; one cache miss at each successive level of the tree. As the CPU-RAM gap grows, this performance degradation will only become worse in the future. We propose a solution that takes advantage of the growing speed of local area networks for clusters. We split the sorted tree structure among the nodes of the cluster. We assume that the structure will fit inside the aggregation of the CPU caches of the entire cluster. We then send a word over the network (as part of a larger packet containing other words) in order to examine the tree structure in another node's CPU cache. We show that this is often faster than the standard solution, which locally incurs multiple cache misses while accessing each successive level of the tree.

cs.DC↗

arXiv2003

DPG: A Cache-Efficient Accelerator for Sorting and for Join Operators

We present a new algorithm for fast record retrieval, distribute-probe-gather, or DPG. DPG has important applications both in sorting and in joins. Current main memory sorting algorithms split their work into three phases: extraction of key-pointer pairs; sorting of the key-pointer pairs; and copying of the original records into the destination array according the sorted key-pointer pairs. The copying in the last phase dominates today's sorting time. Hence, the use of DPG in the third phase provides an accelerator for existing sorting algorithms. DPG also provides two new join methods for foreign key joins: DPG-move join and DPG-sort join. The resulting join methods with DPG are faster because DPG join is cache-efficient and at the same time DPG join avoids the need for sorting or for hashing. The ideas presented for foreign key join can also be extended to faster record pair retrieval for spatial and temporal databases.

cs.DB↗

arXiv2002

Towards a practical, theoretically sound algorithm for random generation in finite groups

This work presents a new, simple O(log^2|G|) algorithm, the Fibonacci cube algorithm, for producing random group elements in black box groups. After the initial O(log^2|G|) group operations, epsilon-uniform random elements are produced using O((log 1/epsilon)log|G|) operations each. This is the first major advance over the ten year old result of Babai [Babai91], which had required O(log^5|G|) group operations. Preliminary experimental results show the Fibonacci cube algorithm to be competitive with the product replacement algorithm. The new result leads to an amusing reversal of the state of affairs for permutation group algorithms. In the past, the fastest random generation for permutation groups was achieved as an application of permutation group membership algorithms and used deep knowledge about permutation representations. The new black box random generation algorithm is also valid for permutation groups, while using no knowledge that is specific to permutation representations. As an application, we demonstrate a new algorithm for permutation group membership that is asymptotically faster than all previously known algorithms.

math.PR↗

arXiv2000

Scalable Parallel Implementation of Geant4 Using Commodity Hardware and Task Oriented Parallel C

We describe a scalable parallelization of Geant4 using commodity hardware in a collaborative effort between the College of Computer Science and the Department of Physics at Northeastern University. The system consists of a Beowulf cluster of 32 Pentium II processors with 128 MBytes of memory each, connected via ATM and fast Ethernet. The bulk of the parallelization is done using TOP-C (Task Oriented Parallel C), software widely used in the computational algebra community. TOP-C provides a flexible and powerful framework for parallel algorithm development, is easy to learn, and is available at no cost. Its task oriented nature allows one to parallelize legacy code while hiding the details of interprocess communications. Applications include fast interactive simulation of computationally intensive processes such as electromagnetic showers. General results motivate wider applications of TOP-C to other simulation problems as well as to pattern recognition in high energy physics.

hep-ph↗