The 2910G is configured via a web browser at its default LAN IP address of 192.168.1.1. There are 15 menus choices along the left hand side of the web configuration GUI. Each menu has three to eleven sub menus, for a total of 80 different configuration screens. Draytek has a nice live demo of all its menu screens located here. I recommend you poke through these screens as you read through the review.
Draytek provides a 214 page manual on the CD that comes with the router. There is a more detailed manual on the Draytek website. Both manuals are concise, but I found them lacking examples and clear explanation on how to configure desired functions.
In the US, Draytek is represented by Strowger Inc. I mention this, as I found Strowger's website to be a useful source of information on networking and configuring the Draytek router, with guides on how to configure VPNs, Load-Balancing, and QoS.
The 2910G supports up to two WAN connections. It has a primary 10/100 Mbps WAN port for connecting a wired ISP service such as a DSL or Cable modem using Static, DHCP, PPPoE, and PPTP/L2TP connection methods. The 2910G also supports a second WAN connection via a second Ethernet or USB-connected WWAN dongle or mobile phone.
Draytek lists 72 different compatible 3G modems / phones on their website. Alternatively, a second wired ISP service such as a DSL or Cable ISP service can be connected to the WAN2/P1 port. When the WAN2/P1 port is enabled as a WAN port, the router will only have 3 LAN ports. I tested the 2910G with two wired Internet connections.
In either dual-WAN configuration, the 2910G supports failover, bandwidth based load-balancing, and protocol specific load-balancing between the two WAN connections. Failover is pretty quick on the 2910G. I set up the router for failover mode with WAN1 as primary, ran a continuous ping to yahoo.com, then disconnected WAN1. As you can see in Figure 4 below, only one ping was lost while the router switched over to the second WAN connection.
Figure 4: Ping test for WAN failover
The default setting is for the 2910G to automatically load balance traffic between WAN1 and WAN2, essentially balancing traffic equally across the two ports. Load balancing can be fine-tuned by defining the upload and download speed of each WAN connection. This is useful if one of the two connections has significantly higher upload or download speeds.
Traffic flows can be defined over the two WAN ports so specific traffic routes primarily over one connection, yet will still failover in case one connection is down. Protocol, source and destination IP addresses, and destination port values can be used to define which traffic flows route over which WAN port.
To test this feature, I set up a simple rule to route ICMP (ping) traffic over the WAN2 interface as shown in Figure 5. While running a continuous ping to yahoo.com, I could see packet counts increasing on the WAN2 interface (via the Online Status menu), but not on the WAN1 interface, verifying the ICMP traffic was indeed routing over the WAN2 interface as specified.
Figure 5: Load balancing policy setup
A possible use for this load balancing feature on the 2910G is to separate UDP delay sensitive traffic from TCP based traffic. For example, UDP based VOIP traffic could be routed over one WAN connection while TCP based web and email traffic could be routed over the other WAN connection, ensuring bandwidth and quality for the VOIP calls and still passing web and email traffic over the other WAN connection.
In addition to bandwidth and protocol based load balancing over Dual-WAN connections, the 2910G can apply bandwidth limits to specific hosts or services, as well as reserve bandwidth for specific applications.
There are three bandwidth configuration options labeled as Session, Bandwidth and Quality of Service. The Session bandwidth configuration option enables limiting bandwidth utilization per connection, such as P2P file transfers. A P2P file transfer is a connection which can consume multiple sessions to complete. By limiting host connections to a maximum number of sessions, the 2910G can control bandwidth consumption by session hungry applications such as BitTorrent.
A more straightforward option is to limit bandwidth utilization by host or IP range using the Bandwidth configuration option. As you can see below, I've restricted host 192.168.1.10 (my laptop) to a 250 kbps upload and download bandwidth limit.
Figure 6: Setting bandwidth limits
To test this configuration, I ran a speed test on my DSL line from my laptop behind the 2910G via speedtest.net with bandwidth limits disabled. My DSL line measured 630 Kbps upload and 11.31 Mbps download. With the bandwidth limit shown in Figure 6 enabled, my DSL line measured 220 Kbps upload and 310 Kbps download, showing the 2910G effectively limited my bandwidth consumption. The 2910G didn't exactly limit bandwidth to 250 Kbps, but it clearly showed its effectiveness.
The third bandwidth configuration option is to implement Quality of Service (QoS). The 2910G allows QoS to be applied to packets via Class Indexes based on IP address, IP Precedence or DSCP values, as well as custom or predefined protocols. There are three customizable Class Indexes, and each can have up to 20 rules of matching conditions. A percentage of the total bandwidth is assigned to each of these three Class Indexes, with the remainder going to all undefined traffic.
For example, I created a simple QoS rule to guarantee 10% of the bandwidth reserved for traffic using the POP3 protocol (Email) and for traffic using the HTTP protocol (Web). Once configured, there is a useful Online Statistics feature that gives a snapshot of bandwidth usage by traffic type as shown in Figure 7.