Search for: All records

The advent of Persistent Memory (PM) devices enables systems to actively persist information at low costs, including program state traditionally in volatile memory. However, this trend poses a reliability challenge in which multiple classes of soft faults that go away after restart in traditional systems turn into hard (recurring) faults in PM systems. In this paper, we first characterize this rising problem with an empirical study of 28 real-world bugs. We analyze how they cause hard faults in PM systems. We then propose Arthas, a tool to effectively recover PM systems from hard faults. Arthas checkpoints PM states via fine-grained versioning and uses program slicing of fault instructions to revert problematic PM states to good versions. We evaluate Arthas on 12 real-world hard faults from five large PM systems. Arthas successfully recovers the systems for all cases while discarding 10× less data on average compared to state-of-the-art checkpoint-rollback solutions. 

The data is from a direct numerical simulation of forced isotropic turbulence on a 4096-cubed periodic grid, using a pseudo-spectral parallel code. The simulations are documented in Ref. 1. Time integration uses second-order Runge-Kutta. The simulation is de-aliased using phase-shifting and truncation. Energy is injected by keeping the energy density in the lowest wavenumber modes prescribed following the approach of Donzis & Yeung. After the simulation has reached a statistical stationary state, a frame of data, which includes the 3 components of the velocity vector and the pressure, are generated and written in files that can be accessed directly by the database (FileDB system). 

The turbulent channel flow database is produced from a direct numerical simulation (DNS) of wall bounded flow with periodic boundary conditions in the longitudinal and transverse directions, and no-slip conditions at the top and bottom walls. In the simulation, the Navier-Stokes equations are solved using a wall {normal, velocity {vorticity formulation. Solutions to the governing equations are provided using a Fourier-Galerkin pseudo-spectral method for the longitudinal and transverse directions and seventh-order Basis-splines (B-splines) collocation method in the wall normal direction. De-aliasing is performed using the 3/2-rule [3]. Temporal integration is performed using a low-storage, third-order Runge-Kutta method. Initially, the flow is driven using a constant volume flux control (imposing a bulk channel mean velocity of U = 1) until stationary conditions are reached. Then the control is changed to a constant applied mean pressure gradient forcing term equivalent to the shear stress resulting from the prior steps. Additional iterations are then performed to further achieve statistical stationarity before outputting fields. 

The data is from a direct numerical simulation (DNS) of homogeneous buoyancy driven turbulence on a 1024-cubed periodic grid. (See README-HBDT.pdf linked document for equations and details.) The simulation was performed with the variable-density version of the petascale CFDNS code. The database covers both the buoyancy driven increase in turbulence intensity as well as the buoyancy mediated turbulence decay. 

The data is from a direct numerical simulation on a 10243 periodic grid of the incompressible MHD equations. (See README-MHD linked document for equations and further details.)