Sang-bum Suh

An OpenCL framework for heterogeneous multicores with local memory

In this paper, we present the design and implementation of an Open Computing Language (OpenCL) framework that targets heterogeneous accelerator multicore architectures with local memory. The architecture consists of a general-purpose processor core and multiple accelerator cores that typically do not have any cache. Each accelerator core, instead, has a small internal local memory. Our OpenCL runtime is based on software-managed caches and coherence protocols that guarantee OpenCL memory consistency to overcome the limited size of the local memory. To boost performance, the runtime relies on three source-code transformation techniques, work-item coalescing, web-based variable expansion and preload-poststore buffering, performed by our OpenCL C source-to-source translator. Work-item coalescing is a procedure to serialize multiple SPMD-like tasks that execute concurrently in the presence of barriers and to sequentially run them on a single accelerator core. It requires the web-based variable expansion technique to allocate local memory for private variables. Preload-poststore buffering is a buffering technique that eliminates the overhead of software cache accesses. Together with work-item coalescing, it has a synergistic effect on boosting performance. We show the effectiveness of our OpenCL framework, evaluating its performance with a system that consists of two Cell BE processors. The experimental result shows that our approach is promising.

Fine-grained I/O access control of the mobile devices based on the Xen architecture

System virtualization is now available for mobile devices allowing for many advantages. Two of the major benefits from virtualization are system fault isolation and security. The isolated driver domain (IDD) model, a widely adopted architecture, enables strong system fault isolation by limiting the impact of driver faults to the driver domain itself. However, excessive I/O requests from a malicious domain to an IDD can cause CPU overuse of the IDD and performance degradation of applications in the IDD and other domains that share the same I/O device with the malicious do-main. If the IDD model is applied to mobile devices, this failure of performance isolation could also lead to battery drain, and thus it introduces a new severe threat to mobile devices. In order to solve this problem, we propose a fine-grained I/O access control mechanism in an IDD. Requests from guest domains are managed by an accounting module in terms of CPU usage, with the calcula-tion of estimated CPU consumption using regression equations. The requests are scheduled by an I/O access control enforcer ac-cording to security policies. As a result, our mechanism provides precise control on the CPU usage of a guest domain due to I/O device access, and prevents malicious guest domains from CPU overuse, performance degradation, and battery drain. We have implemented a prototype of our approach considering both network and storage devices with a real smart phone (SGH-i780) that runs two para-virtualized Linux kernels on top of Secure Xen on ARM. The evaluation shows our approach effectively protects a smart phone against excessive I/O attacks and guarantees availability.

Chameleon: a capability adaptation system for interface virtualization

As computing capabilities of embedded devices increase, services such as Internet and multimedia are comportably supported by such devices. Users may demand for migrating services from one machine to another that can provide better capabilities and convenience in terms of battery stock and input/output (I/O) method. However, I/O devices capabilities between embedded mobile devices and user interaction behavior are particulary different. So, heterogenous I/O capabilities and functions may make migration of computing states between embedded devices virtually impossible. To guarantee seamless I/O state migration between embedded mobile devices, we propose interface virtualization and capability adaptation. In this paper, we describe a system architecture based on Xen virtualization as well as design principles and performance.

Fine-grained I/O access control based on Xen virtualization for 3G/4G mobile devices

Although Xens isolated driver domain (IDD) model enables strong system isolation by limiting the impact of driver faults to the driver domain itself, it results in severe security problems when malware in a guest domain tries to abuse mobile devices limited system resources by sending an extreme number of I/O requests to the IDD. In order to solve this problem, this paper presents a fine-grained I/O access control mechanism in an IDD. Requests from guest domains are managed by an accounting module in terms of CPU usage, with the calculation of estimated CPU consumption using regression equations. The requests are scheduled by an I/O access control enforcer according to security policies. As a result, our mechanism provides precise control on the CPU usage of a guest domain due to I/O device access, and prevents compromised guest domains from CPU overuse, performance degradation, and battery drain. We have implemented a prototype of our approach considering both network and storage devices with a real smart phone (SGH-i780) that runs two para-virtualized Linux kernels on top of Secure Xen on ARM. The evaluation shows our approach effectively protects a smart phone against excessive I/O attacks and guarantees availability.

Enhancing init scheme for improving bootup time in mobile devices

The init process has evolved as a system and service manager of process scheduling, input/output scheduling, memory scheduling, and the creation and termination of processes. The addition of software functionality increases software complexity. In a mobile environment, the increase of software complexity results in a slow boot time for mobile devices. This paper describes a mobile aware bootup framework by isolating the dependencies of the execution sequence among the processes, and by simplifying the existing init scheme to accelerate the boot time of the mobile device. From our experiment, our proposed approaches reduce the bootup time of the mobile device by about 50%.

User-level memory scheduler for optimizing application performance in NUMA-based multicore systems

Multicore CPU architectures have been established as a structure for general-purpose systems for high-performance processing of applications. Recent multicore CPU has evolved as a system architecture based on non-uniform memory architecture. For the technique of using the kernel space that shifts the tasks to the ideal memory node, the characteristics of the applications of the user-space cannot be considered. Therefore, kernel level approaches cannot execute memory scheduling to recognize the importance of user applications. Moreover, users need to run applications after sufficiently understanding the multicore CPU based on non-uniform memory architecture to ensure the high performance of the users applications. This paper presents a user-space memory scheduler that allocates the ideal memory node for tasks by monitoring the characteristics of non-uniform memory architecture. From our experiment, the proposed system improved the performance of the application by up to 25% compared to the existing system.

Preemptibility-aware responsive multi-core scheduling

We propose a novel responsive scheduling technique to minimize the scheduling latency of a real-time process in the multi-core architecture, called the preemptibility-aware scheduling (PAS). Modern complex operating systems contain numerous long interrupt-disabled and non-preemptible sections, and consequently these sections obstruct the immediate handling of urgent interrupts and the rapid scheduling of interrupt-driven real-time tasks, causing significant latency between interrupt arrival and process scheduling. The proposed PAS guarantees that before an urgent interrupt occurs, at least one among multiple CPU cores is always in both interrupt-enabled and preemptible sections, so that the incoming urgent interrupt can always be handled with no significant delay by such CPU core. Experimental results show that the worst-case latency can be reduced by 27-96%.

A Virtual Window System for CE Devices Based on System Virtualization

Owing to the benefits of system virtualization, even CE devices have come to take advantage of the technology. However, due to the lack of windowing system which fits virtualization-based CE devices, it is not only inefficient but also difficult for end users to utilize the CE devices running multiple domains. In this paper we present an effective virtual window system for CE devices based on system virtualization. Our approach has three major advantages: (1) by modifying X window system, it provides shared windowing services between domains without dependency of a specific network protocol; (2) by providing a unified graphical user interface which integrates icons of all the applications from every domain, it frees users from remembering which applications are located in which domain and from doing tedious operations for application launching and installation; (3) it provides efficiency in terms of size (storage and memory) and performance to CE devices. We have implemented a prototype of the virtual window system on the basis of Secure Xen on ARM. Our evaluation shows that our approach is usable and efficient enough to be practically adopted for CE devices based on system virtualization.

CloudSwap: A Cloud-Assisted Swap Mechanism for Mobile Devices

Application caching is a key feature to enable fast application switches for mobile devices by caching the entire memory pages of applications in the devices physical memory. However, application caching requires a prohibitive amount of memory unless a swap feature is employed to maintain only the working sets of the applications in memory. Unfortunately, mobile devices often disable the invaluable swap feature as it can severely decrease the flash-based local storage devices already marginal lifespan due to the increased writes to the device. As a result, modern mobile devices suffering from the insufficient memory space end up killing memory-hungry applications and keeping only a few applications in the memory. In this paper, we propose CloudSwap, a fast and robust swap mechanism for mobile devices to enable the memory-oblivious application caching. The key idea of CloudSwap is to use the fast local storage as a cache of read-intensive swap pages, while storing prefetch-enabled, write-intensive swap pages in a cloud storage. To preserve the lifespan of the local storage, CloudSwap minimizes the number of writes to the local storage by storing the modified portions of the locally swapped pages in a cloud. To reduce the remote swap-in latency, CloudSwap exploits two cloud-assisted prefetch schemes, the app-aware read-ahead scheme and the access pattern-aware prefetch scheme. Our evaluation shows that the performance of CloudSwap is comparable to a fast, but lifespan-critical local swap system, with only 18% lifespan reduction, compared to the local swap systems 85% lifespan reduction.

Raptor codes-based energy-efficient multipath multimedia transport protocol over HetNets

In this work, we propose a Raptor codes-based energy-efficient multipath multimedia transport protocol over heterogeneous wireless networks (HetNets) in order to provide seamless high-quality video streaming service in an energy efficient way. In the proposed protocol, systematic Raptor codes are employed to mitigate the video quality degradation. For energy saving, the proposed protocol not only shapes a video stream into bursts, but also determines the Raptor encoding parameters including code rate, symbol size, and the number of source symbols for energy saving. The proposed protocol is fully implemented on an Android-based smartphone, and examined over real HetNets.