Panlab: Brewing IDEAS

Making home Internet faster has little to do with “speed”

04/18/2023 Mike Conlow 8 min read This post is also available in 简体中文, 繁體中文, 日本語, 한국어, Français, Deutsch and Español.

More than ten years ago, researchers at Google published a paper with the seemingly heretical title “More Bandwidth Doesn’t Matter (much)”. We published our own blog showing it is faster to fly 1TB of data from San Francisco to London than it is to upload it on a 100 Mbps connection. Unfortunately, things haven’t changed much. When you make purchasing decisions about home Internet plans, you probably consider the bandwidth of the connection when evaluating Internet performance. More bandwidth is faster speed, or so the marketing goes. In this post, we’ll use real-world data to show both bandwidth and – spoiler alert! – latency impact the speed of an Internet connection. By the end, we think you’ll understand why Cloudflare is so laser focused on reducing latency everywhere we can find it.

The grand summary of the blog that follows is this:
* There are many ways to evaluate network performance.
* Performance “goodness” depends on the application -- a good number for one application can be of zero benefit to a different application.
* “Speed” numbers can be misleading, not least because any single metric cannot accurately describe how all applications will perform.

juvana wrote:Network support perspectives encompass a range of crucial roles in ensuring seamless connectivity and system reliability. Network administrators oversee infrastructure, ensuring its stability and security. business internet providers near me.Helpdesk technicians troubleshoot issues for end-users, enhancing their experience. Network engineers design, deploy, and optimize network solutions. Together, they form the backbone of efficient data transmission, enabling businesses to thrive in our increasingly interconnected world.

as an appreciation token to my postdoc supervisor as well

Chapter 7

Teaching from Home: Computer and Communication Network Perspectives

Jianping Pan

which can be useful for

family Skype video calls

too

fei’s setup

poster

extra: zoom teaching tips

4.3 Recommendations on Teaching from Home

Based on the above summary, comparison and proposal, and the experience in Spring and Summer 2020, in this section we make some recommendations on online teaching in Fall 2020 and beyond. First, use a computer with Ethernet connection to home router whenever possible, and choose an ISP with reasonable data rates, especially the uplink one, but more importantly with less delay and variation and fewer packet losses and service outages. When wired Ethernet is not available, consider NNW or improved WiFi with wired or wireless interconnection if needed. When feasible and affordable, consider to have two independent ISPs to guarantee the reliability for teaching from home, especially when large-scale synchronous lecturing is anticipated. If there are other active users at home at the same time, consider to allocate them to use a low-priority WiFi channel and ISP when possible to avoid link congestion.

after we addressed the technical issues for teaching/learning/working from home/dorm/anywhere, we can return to pedagogical ones

4.2.3 Load Balancing

Beyond primary and backup ISPs, it is also possible to bond both DSL and CM ISPs at the same time, through a technique known as load balancing, i.e., some connections use one ISP and others use another, either equally or proportionally to a predefined or self-learned weight, as shown in Figure 1. The advantage is obvious: user can utilize both links if paid already, and each can backup the other for reliability. However, it requires more sophisticated configuration at the home gateway, where two upstream default routes have to be maintained at the same time, one for each group of flows. Open-source routers such as those powered by OpenWRT and DD-WRT have user-contributed scripts to automatically create virtual LAN (VLAN) for different upstream ISPs, define rules to split traffic, check network connectivity periodically, and fail over to the other link when necessary, under the so-called “DualWAN” capability. Most full-blown Linux systems, e.g., Ubuntu, have multi-homing capability, and some low-cost SMB routers, such as TP-Link R470T+, offer multi-WAN capability with very simple and intuitive graphic user interface (GUI)-based configuration. Table 2 lists the delay and throughput to speedtest servers hosted by Cable and DSL ISP, through Cable and DSL individually, and jointly as bonded. It shows the great advantage of bonding.

However, there are still some subtle issues with load balancing in terms of the “bonding” granularity, i.e., whether the packets from the same session can be distributed over different upstream ISPs. If so, a single application can fully benefit from both ISPs, in terms of both reliability and capacity, but this capability depends on specific applications and whether they or the transport-layer protocol they use can deal with out-of-order packet arrivals through different paths. For most CLM tools, even free but not open source, we cannot guarantee their behavior. Nevertheless, they seem to be able to handle when video and audio streams are carried by different ISPs, similar in concept but different in technology as the call-in feature in most commercial CLM tools. Table 3 compares the interruption due to host interface and Internet access down and up events for App and Web-based CLM platforms. With bonding, load balancing and liveliness checking, CLM only suffers in the order of the detection timer, which can be as low as 1 sec and much lower than the down-to-up time of DSL (40 sec) and CM (few minutes).

Table 3: CLM interruption: host vs Internet link down vs up

native app has better performance https://openwrt.org/docs/guide-user/network/wan/

WAN (aka Internet access)

https://wiki.dd-wrt.com/wiki/index.php/ ... y:Dual_WAN

Dual WAN

https://www.amazon.com/slp/load-balanci ... ck437yc8ov

Load Balancing Router

4.2.2 Primary vs Backup

As shown in Figure 1, we subscribed to two ISPs, one DSL and one CM, which are often available in and competing for the same market. Note that some DSL and CM ISPs wholesale from other major ISPs and then resale to consumers, but here we know these two ISPs are actually independent in terms of their wiring infrastructures and maintenance schedules, to improve reliability. Depending on the service quality and cost of these two ISPs, one can be designated as the primary upstream ISP (e.g., the one offers a flat monthly fee or without data cap) and the other backup (the one charges by the data amount transferred, including cellular or satellite ISPs). To facilitate the automatic switch between the primary and backup upstream ISP, the WiFi AP (or an interconnected group of them) with routing functionalities and connected to both DSL and cable modems shall check the liveliness of the primary ISP, e.g., by pinging a known IP address periodically, and then set the default route to the backup ISP when the primary one fails. Depending on the user-defined policy, the home gateway can keep checking the primary ISP periodically and switch back when the primary one becomes available. In this case, there is only one active ISP at any time by default routing. It improves the reliability, unless both fail at the same time, without additional capacity.

Most modern Web-era applications, including Blackboard, WebEx and Zoom, can sustain the switch of ISPs, and thus the change of the publicly assigned IP address, during an active audio and video session, as these applications keep their session states and recognize mobile users in the application layer (e.g., by HTTP cookies), instead of by IP addresses and TCP or UDP port numbers. When one connection fails, others are automatically created to continue the session, similar to multi-path TCP (MPTCP). This is also used by many smart phones to switch between WiFi and cellular connections automatically. For old, single-connection applications such as ssh, however, users have to reconnect manually.

Table 2: Individual and bonded speed test: Cable vs DSL ISP

https://en.wikipedia.org/wiki/Multipath_TCP

Multipath TCP (MPTCP) is an ongoing effort of the Internet Engineering Task Force's (IETF) Multipath TCP working group, that aims at allowing a Transmission Control Protocol (TCP) connection to use multiple paths to maximize resource usage and increase redundancy.

https://www.multipath-tcp.org/

Welcome to the Linux Kernel MultiPath TCP project

https://www.ietfjournal.org/multipath-tcp-deployments/

Multipath TCP Deployments

https://multipath-quic.org/

Multipath QUIC

4.2 WAN Reliability Augmentation

Both consumer-grade DSL and cable Internet access services suffer reliability issues, far below what fiber optics can offer in commercial workplace. For instructors to lose connection to the Internet, even briefly or intermittently, is unacceptable for a potentially large group of students during lectures. In the following, we examine and compare DSL and CM-based Internet access, and the possibility to leverage both ISPs when feasible to improve reliability.

4.2.1 DSL vs Cable Modem

As discussed in Section 3.3.2 and 3.3.3, DSL and cable both have their pros and cons. DSL is not affected by neighbors but has limited bandwidth and is more susceptible to noise and interference. CM has more bandwidth but has to share the capacity with neighbors, especially for the uplink. E.g., an advertised 25/5Mbps (for down and uplink, respectively) DSL plan only achieves a 3Mbps uplink, but the ping time from the DSL modem to the first DSL ISP router is lower and more stable due to the dedicated uplink. An advertised 50/5Mbps CM plan can achieve a 59Mbps downlink during off-peak hours, but its ping time to the first CM ISP router is a bit higher and highly variable due to the shared capacity, as shown in Figure 3. According to most CLM platforms, a 500kbps uplink is sufficient for a standard-definition video stream, which is well accommodated by most DSL and CMlinks, but delay and loss affect the live video streaming much more.

However, from the DSL and CM ISP networks to CLM data centers, depending on how and where CLM providers deploy their services, the varying bandwidth and delay can cause additional QoS fluctuation, as illustrated in Table 1 with traceroute to a public enhanced DNS server. In terms of reliability, both DSL and CM can vary by providers and regions, the cable plant and supporting infrastructures. Consumer-grade ISPs and plans also have routine maintenance and unexpected outage without guaranteed backup and recovery allowed by their service agreement. Thus, relying on one DSL or CM service provider is often not sufficient for high reliability. Paying higher cost for a business service plan is an option, but in the following we explore other more flexible alternatives.

Table 1: Traceroute from home to 1.1.1.1: Cable vs DSL ISP

your mileage may vary with your service providers on what they advertise vs deliver