Michio Honda

mSwitch: a highly-scalable, modular software switch

In recent years software network switches have regained eminence as a result of a number of growing trends, including the prominence of software-defined networks, as well as their use as back-ends to virtualization technologies, to name a few. Consequently, a number of high performance switches have been recently proposed in the literature, though none of these simultaneously provide (1) high packet rates, (2) high throughput, (3) low CPU usage, (4) high port density and (5) a flexible data plane. This is not by chance: these features conflict, and while achieving one or a few of them is (now) a solved problem, addressing the combination requires significant new design effort. In this paper we fill the gap by presenting mSwitch. To prove the flexibility and performance of our approach, we use mSwitch to build four distinct modules: a learning bridge consisting of 45 lines of code that outperforms FreeBSDs bridge by up to 8 times; an accelerated Open vSwitch module requiring small changes to the code and boosting performance by 2.6--3 times; a protocol demultiplexer for userspace protocol stacks; and a filtering module that can direct packets to virtualized middleboxes.

Rekindling network protocol innovation with user-level stacks

Recent studies show that more than 86% of Internet paths allow well-designed TCP extensions, meaning that it is still possible to deploy transport layer improvements despite the existence of middleboxes in the network. Hence, the blame for the slow evolution of protocols (with extensions taking many years to nbecome widely used) should be placed on end systems. In this paper, we revisit the case for moving protocols stacks up into user space in order to ease the deployment of new protocols, extensions, or performance optimizations. We present MultiStack, operating system support for user-level protocol stacks. MultiStack runs within commodity operating systems, can concurrently host a large number of isolated stacks, has a fall-back path to the legacy host stack, and is able to process packets at rates of 10Gb/s. We validate our design by showing that our mux/demux layer can validate and switch packets at line rate (up to 14.88 Mpps) on a 10 Gbit port using 1-2 cores, and that a proof-of-concept HTTP server running over a basic userspace TCP outperforms by 18-90% both the same server and nginx running over the kernels stack.

Unikernels Everywhere: The Case for Elastic CDNs

Video streaming dominates the Internets overall traffic mix, with reports stating that it will constitute 90% of all consumer traffic by 2019. Most of this video is delivered by Content Delivery Networks (CDNs), and, while they optimize QoE metrics such as buffering ratio and start-up time, no single CDN provides optimal performance. In this paper we make the case for elastic CDNs, the ability to build virtual CDNs on-the-fly on top of shared, third-party infrastructure at a scale. To bring this idea closer to reality we begin by large-scale simulations to quantify the effects that elastic CDNs would have if deployed, and build and evaluate MiniCache, a specialized, minimalistic virtualized content cache that runs on the Xen hypervisor. MiniCache is able to serve content at rates of up to 32 Gb/s and handle up to 600K reqs/sec on a single CPU core, as well as boot in about 90 milliseconds on x86 and around 370 milliseconds on ARM32.

Designing a Resource Pooling Transport Protocol

This paper presents a design for an end-to-end transport protocol for multi-homed end systems that pools the communication resources of multiple network paths to support a single communication session. This approach offers improved performance and resilience compared to communicating over a single path. Compared to previous efforts, deployability in the current commercial Internet, i.e., in the presence of middleboxes, filtering and restricted connectivity, was a key driver for the design of the Resource Pooling Protocol (RPP).

iPath: Achieving high-performance end-to-end paths for multi-homed mobile hosts

This work aims at building iPath - a path-selection system for multi-homed mobile hosts that assists vertical handoff. Unlike existing systems, iPath selects the access network interface for communications based on end-to-end path properties (e.g., available bandwidth, delay, packet loss rate, and jitter). iPath also adopts a new metric for handoff - “switching cost”, which is caused by the congestion control behavior of transport protocols when the mobile host changes its source or destination address. Handoff decision based on the switching cost prevents inappropriate handoff in terms of transmission performance and smoothness.

A Connectivity-Driven Retransmission Scheme Based On Transport Layer Readdressing

Migration between different wireless access networks often involves disconnected period, which is caused by passing an area of bad wireless coverage and potential overhead to switch the network on the network interface to connect to. The disconnected period can cause extra transmission delay due to the timer-driven retransmission behavior in the transport protocols, such as TCP and SCTP. We propose a new retransmission scheme to achieve better migration performance in SCTP, which is a newer connection-oriented and reliable transport protocol that is becoming popular. Our scheme minimizes the extra transmission delay by leveraging address reconfiguration operation in SCTP without involving other layers. It decreases the delay more than 5 seconds compared to the original SCTP when migration involves approximately ten-second disconnected period. The implementation of our scheme is already imported in FreeBSD.

SmSCTP: A Fast Transport Layer Handover Method Using Single Wireless Interface

Many handover techniques in the Internet have been introduced with the development of mobile computing technologies. Although many proposed handover schemes utilize multiple interfaces, having multiple interfaces can increase power consumption, device installation space and hardware costs of mobile devices. Therefore, we have been studying handover schemes for mobile nodes with a single wireless network interface. To achieve seamless and efficient handover, we focus on stream control transmission protocol (SCTP) which offers message-oriented, reliable and connection-oriented delivery transport service. Unlike other transport protocols such as TCP, SCTP can provide an end-to-end handover mechanism with a multi-homing feature. However, the handover mechanism in the current SCTP causes large handover delay especially when a mobile node has only one single wireless network interface. In this paper, we have investigated the current issues in the SCTP handover mechanism and propose a new handover scheme based on SCTP, which identifies a communication path as a pair of source and destination address. We also propose new data retransmission feature for smooth handover.

HyperNF: building a high performance, high utilization and fair NFV platform

Network Function Virtualization has been touted as the silver bullet for tackling a number of operator problems, including vendor lock-in, fast deployment of new functionality, converged management, and lower expenditure since packet processing runs on inexpensive commodity servers. The reality, however, is that, in practice, it has proved hard to achieve the stable, predictable performance provided by hardware middleboxes, and so operators have essentially resorted to throwing money at the problem, deploying highly underutilized servers (e.g., one NF per CPU core) in order to guarantee high performance during peak periods and meet SLAs. In this work we introduce HyperNF, a high performance NFV framework aimed at maximizing server performance when concurrently running large numbers of NFs. To achieve this, HyperNF implements hypercall-based virtual I/O, placing packet forwarding logic inside the hypervisor to significantly reduce I/O synchronization overheads. HyperNF improves throughput by 10%-73% depending on the NF, is able to closely match resource allocation specifications (with deviations of only 3.5%), and to efficiently cope with changing traffic loads.

PASTE: Network Stacks Must Integrate with NVMM Abstractions

This paper argues that the lack of explicit support for non-volatile main memory (NVMM) in network stacks fundamentally limits application performance. NVMM devices have been integrated into general-purpose OSes by providing familiar file-based interfaces and efficient byte-granularity access by bypassing page caches. However, this powerful property cannot be fully utilized unless network stacks also support it and applications exploit such support. This requires a thoroughly new network stack design, including low-level buffer management and APIs. We propose such a new network stack architecture to support NVMM and demonstrate its advantages for efficient write-ahead logging, a popular technique to implement transactions.

PACUE: processor allocator considering user experience

GPU accelerated applications including GPGPU ones are commonly seen in modern PCs. If many applications compete on the same GPU, the performance will decrease significantly. Some applications have a large impact on user experience. Therefore, for such applications, we have to limit GPU utilization by the other applications. It might be straightforward to modify applications to switch compute device dynamically for intelligent resources allocation. Unfortunately, we cannot do so due to software distribution policy or the other reasons. In this paper, we propose PACUE, which allows the end system to allocate compute devices arbitrary to applications. In addition, PACUE guesses optimal compute device for each application according to user preference. We implemented the dynamic compute device redirector of PACUE including OpenCL API hooking and device camouflaging features. We also implemented the frame of the resource manager of PACUE. We demonstrate PACUE achieves dynamic compute device redirecting on one out of two real applications and on all of 20 sample codes.