Ada Gavrilovska

GViM: GPU-accelerated virtual machines

The use of virtualization to abstract underlying hardware can aid in sharing such resources and in efficiently managing their use by high performance applications. Unfortunately, virtualization also prevents efficient access to accelerators, such as Graphics Processing Units (GPUs), that have become critical components in the design and architecture of HPC systems. Supporting General Purpose computing on GPUs (GPGPU) with accelerators from different vendors presents significant challenges due to proprietary programming models, heterogeneity, and the need to share accelerator resources between different Virtual Machines (VMs). To address this problem, this paper presents GViM, a system designed for virtualizing and managing the resources of a general purpose system accelerated by graphics processors. Using the NVIDIA GPU as an example, we discuss how such accelerators can be virtualized without additional hardware support and describe the basic extensions needed for resource management. Our evaluation with a Xen-based implementation of GViM demonstrate efficiency and flexibility in system usage coupled with only small performance penalties for the virtualized vs. non-virtualized solutions.

VM power metering: feasibility and challenges

This paper explores the feasibility of and challenges in developing methods for black-box monitoring of the power usage of a virtual machine (VM) at run-time, on shared virtualized compute platforms, including those with complex memory hierarchies. We demonstrate that VM-level power utilization can be accurately estimated, or estimated with accuracy with bound error margins. The use of bounds permits more lightweight online monitoring of fewer events, while relaxing the fidelity of the estimates in a controlled manner. Our methodology is evaluated on the Intel Core i7 and Core2 x86-64 platforms, running synthetic and SPEC benchmarks.

Differential virtual time (DVT): rethinking I/O service differentiation for virtual machines

This paper investigates what it entails to provide I/O service differentiation and performance isolation for virtual machines on individual multicore nodes in cloud platforms. Sharing I/O between VMs is fundamentally different from sharing I/O between processes because guest VM operating systems use adaptive resource management mechanisms like TCP congestion avoidance, disk I/O schedulers, etc. The problem is that these mechanisms are generally sensitive to the magnitude and rate of change of service latencies, where failing to address these latency concerns while designing a service differentiation framework for I/O results in undue performance degradation and hence, insufficient isolation between VMs. This problem is addressed by the notion of Differential Virtual Time (DVT), which can provide service differentiation with performance isolation for VM guest OS resource management mechanisms. DVT is realized within a proportional share I/O scheduling framework for the Xen hypervisor, and its use requires no changes to guest OSs. DVT is applied to message-based I/O, but is also applicable to subsystems like disk I/O. Experimental results with DVT-based I/O scheduling for representative applications demonstrate the utility and effectiveness of the approach.

S-NFV: Securing NFV states by using SGX

Network Function Virtualization (NFV) applications are stateful. For example, a Content Distribution Network (CDN) caches web contents from remote servers and serves them to clients. Similarly, an Intrusion Detection System (IDS) and an Intrusion Prevention System (IPS) have both per-flow and multi-flow (shared) states to properly react to intrusions. On todays NFV infrastructures, security vulnerabilities many allow attackers to steal and manipulate the internal states of NFV applications that share a physical resource. In this paper, we propose a new protection scheme, S-NFV that incorporates Intel Software Guard Extensions (Intel SGX) to securely isolate the states of NFV applications.

pVM: persistent virtual memory for efficient capacity scaling and object storage

Next-generation byte-addressable nonvolatile memories (NVMs), such as phase change memory (PCM) and Memristors, promise fast data storage, and more importantly, address DRAM scalability issues. State-of-the-art OS mechanisms for NVMs have focused on improving the block-based virtual file system (VFS) to manage both persistence and the memory capacity scaling needs of applications. However, using the VFS for capacity scaling has several limitations, such as the lack of automatic memory capacity scaling across DRAM and NVM, inefficient use of the processor cache and TLB, and high page access costs. These limitations reduce application performance and also impact applications that use NVM for persistent object storage with flat namespaces, such as photo stores, NoSQL databases, and others. To address such limitations, we propose persistent virtual memory (pVM), a system software abstraction that provides applications with (1) automatic OS-level memory capacity scaling, (2) flexible memory placement policies across NVM, and (3) fast object storage. pVM extends the OS virtual memory (VM) instead of building on the VFS and abstracts NVM as a NUMA node with support for NVM-based memory placement mechanisms. pVM inherits benefits from the cache and TLB-efficient VM subsystem and augments these further by distinguishing between persistent and nonpersistent capacity use of NVM. Additionally, pVM achieves fast persistent storage by further extending the VM subsystem with consistent and durable OS-level persistent metadata. Our evaluation of pVM with memory capacity-intensive applications shows a 2.5x speedup and up to 80% lower TLB and cache misses compared to VFS-based systems. pVMs object store provides 2x higher throughput compared to the block-based approach of the state-of-the art solution and up to a 4x reduction in the time spent in the OS.

Service morphing: integrated system- and application-level service adaptation in autonomic systems

Service morphing is a set of techniques used to continuously meet an applications quality of service (QoS) needs, in the presence of run-time variations in service locations, platform capabilities, or end-user needs. These techniques provide high levels of flexibility in how, when, and where necessary processing and communication actions are performed. Lightweight middleware supports flexibility by permitting end-users to subscribe to information channels of interest to them whenever they desire, and then apply exactly the processing to such information they require. New compiler and binary code generation techniques dynamically generate, deploy, and specialize code in order to match current user needs to available platform resources. Finally, to deal with run-time changes in resource availability, kernel-level resource management mechanisms are associated with user-level middleware. Such associations range from loosely coupled, where kernel-level resource management monitors and occasionally responds to userlevel events, to tightly coupled, where kernel-level mechanisms import, export, and use performance and control attributes in conjunction with each resource-relevant userlevel event.

IBMon: monitoring VMM-bypass capable InfiniBand devices using memory introspection

Active monitoring of virtual machine (VM) behaviors and their utilization of different resource types is critical for effective management of high end cluster machines, data centers, and cloud computing infrastructures. Unfortunately, for reasons of performance, certain types of existing and future devices support capabilities that provide VMs with direct access to device resources, thereby bypassing the virtual machine monitor (VMM). This presents significant challenges to the VMM due to its resulting inability to assess VMdevice interactions. This paper describes a monitoring utility, IBMon, which enables the asynchronous monitoring of virtualized Infini-Band devices -- a sample VMM-bypass device heavily used in the HPC community. In the absence of adequate hardware supported monitoring information, IBMon uses dynamic memory introspection techniques to infer information regarding the VM-device interactions. Experimental results demonstrate that IBMon can asynchronously monitor VMM-bypass operations with acceptable accuracy, and negligible overheads, including for larger number of VMs, and for VMs with dynamic behavior patterns.

ECC: Edge Cloud Composites

With an ever increasing number of networked devices used in mobile settings, or residing in homes, offices, and elsewhere, there is a plethora of potential computational infrastructure available for providing end users with new functionality and improved experiences for their interactions with the cyber physical world. The goal of our research is to permit end user applications to take advantage of dynamically available, local and remote computational infrastructure, without requiring applications to be explicitly rewritten and/or reconfigured for each scenario and with minimal end user intervention. Edge Cloud Composites (ECC) make possible the dynamic creation of virtual computational platforms that (i) can be composed from specific capabilities - competences - of participating devices, (ii) are guided by end user-centric abstractions capturing current user context and user intent, and (iii) use dynamic methods for device discovery and ECC maintenance. In contrast to datacenter clouds, ECC participants can include both virtualized and non-virtualized devices, and in addition, services running remotely, made possible by ECCs CIC abstractions, where C(ompetence) captures the functional capabilities of accessible devices and/or remote services, (I)ntent articulates end user desires, and (C)ontext describing the current operating environment. Concrete examples prototyped in this work include Android applications for distributed video playback, collaborative UI, and a distributed augmented reality application. For all such applications, an ECC composed from available devices, and guided by ECCs CIC notions, obtains up to 86% performance improvements and reductions in energy consumption of up to 37% compared to running on a single device. A resultant advantage in using ECCs to run applications is the ability to avoid the unpredictable latency variations seen in device-remote cloud interactions.

Spatially-Aware Optimization of Energy Consumption in Consolidated Data Center Systems

Energy efficiency in data center operation depends on many factors, including power distribution, thermal load and consequent cooling costs, and IT management in terms of how and where IT load is placed and moved under changing request loads. Current methods provided by vendors consolidate IT loads onto the smallest number of machines needed to meet application requirements. This paper’s goal is to gain further improvements in energy efficiency by also making such methods ’spatially aware’, so that load is placed onto machines in ways that respect the efficiency of both cooling and power usage, across and within racks. To help implement spatially aware load placement, we propose a model-based reinforcement learning method to learn and then predict the thermal distribution of different placements for incoming workloads. The method is trained with actual data captured in a fully instrumented data center facility. Experimental results showing notable differences in total power consumption for representative application loads indicate the utility of a two-level spatially-aware workload management (SpAWM) technique in which (i) load is distributed across racks in ways that recognize differences in cooling efficiencies and (ii) within racks, load is distributed so as to take into account cooling effectiveness due to local air flow. The technique is being im

Fast, Scalable and Secure Onloading of Edge Functions Using AirBox

This paper argues for the utility of back-end driven onloading to the edge as a way to address bandwidth use and latency challenges for future device-cloud interactions. Supporting such edge functions (EFs) requires solutions that can provide (i) fast and scalable EF provisioning and (ii) strong guarantees for the integrity of the EF execution and confidentiality of the state stored at the edge. In response to these goals, we (i) present a detailed design space exploration of the current technologies that can be leveraged in the design of edge function platforms (EFPs), (ii) develop a solution to address security concerns of EFs that leverages emerging hardware support for OS agnostic trusted execution environments such as Intel SGX enclaves, and (iii) propose and evaluate AirBox, a platform for fast, scalable and secure onloading of edge functions.