Most people never consider the throughput or the power efficiency of their home router. Pretty much any router on the market is more than capable enough for today's broadband connections. The average home user has a single IP address that is dynamically assigned by their ISP. Many will use an inexpensive router to provide internet access to multiple computers within their home. Some power users and SOHO users have one or more static IP addresses assigned by their ISP. It is fairly common for people to run a web server or other type of server on one or more of the public IP address and have their router on another. Data from the local internal network to the server runs through the router. This report discusses a test procedure for measuring the throughput of a home router then measures the performance of several commercially available routers as well as some PC based routers. Special attention is given to power efficiency
This investigation started when it took over an hour to transfer a backup of a mail server on the public side of an inexpensive router to a laptop on the local area network. Rough calculations using the size of the file and time to transfer showed the data rate to be around 5 megabits per second. This seemed a bit low for a 100 megabit network. Hence this investigation and this report.
All of the devices tested for this report have the minimal required function set of routing between a WAN and a LAN, providing Network Address Translation (NAT) and assigning IP addresses to devices on the LAN using DHCP. Most support a variety of authentications to ISP's on the WAN, offer a variety of routing rules, stateful packet inspection (SPI) and provide web based administration. A few offer VPN capability and traffic shaping. All of these are useful features and their use is covered elsewhere. This report focuses on the throughput of the various devices and their power efficiency.
A very brief survey of network performance measurement tools concluded with the selection of iperf from dast.nlanr.net to measure the raw throughput of a TCP/IP network. These tests only considered total throughput with full packets. No attempt was made to determine the routers packet per second limit. One handy Linux machine was selected as the iperf server and a laptop was selected as the client. Both the Linux and Windows versions of the client were tested with no measurable difference between the two. Tests were run with the laptop connected to the same switch as the public machine and with various routers between the laptop and the public machine.
The one thing in this universe that is ultimately limited is energy. Putting aside the eventual entropy that will engulf the universe in trillions of years, we can look at the short term benefits of energy efficient equipment in our monthly electricity bill. A kill-a-watt meter was used to measure the power drawn from the power plug for the routers tested, both idle and under full load. This can be used to estimate the cost to operate the router 24 hours a day, 7 days a week. The ratio of throughput to power consumed is used as a measure of power efficiency.
Raw data for the systems measured follows.
|Direct connection||95 Mbit/sec|
|Linksys befsr41||5W||4.5 Mbit/sec||0.9 Mb/W|
|Soekris net4501 with m0n0wall||6W||17.1 Mbit/sec||2.8 Mb/W|
|Linksys wrt54g||6W||27 Mbit/sec||4.5 Mb/W|
|Intel ISP1100 with 600 MHz Celeron running Linux||40w||78 Mbit/sec||1.9 Mb/W|
|Rooter Router||26W||94 Mbit/sec||3.6 Mb/W|
|D-Link DI-701 Residential Gateway||4W||7.9 Mbit/sec||2.0 Mb/W|
|Wrap with m0n0wall||6W||38 Mbit/sec||6.3 Mb/W|
|Router board 532A with Routeros||4W||52.6 Mbit/sec||13.2 Mb/W|
|Cisco 1605||8W||8.2 Mbit/sec||1.0 Mb/W|
|Caymen 2E Model 500-H||10W||8.0 Mbit/sec||0.8 Mb/W|
All of off the shelf SOHO routers tested were more than capable of providing basic routing and firewall capabilities for the average home user. The slowest might be a bottleneck for people with high speed (6 mbit/second) cable. Folk with high speed broadband connections might consider one of the faster SOHO routers.
Creating your own firewall router from existing "hand me down" hardware can be a satisfying experience with lots to learn or an exercise in futility depending on how one chooses to view it. Power efficiency and reliability may be a concern using hand me down equipment. Learning to write your own firewall rules is a learning experience that can be avoided by using software packages to hide the rules. The careful selection of hardware selected for power efficiency along with software designed specifically for routing can provide a variety of performance metrics at different price points to meet the needs of most.
Last modified: July 18 2006 20:27:59
Copyright © 2001-2016 by Fractal
Hosting by ziandra.net