Speech:Network


 * Home
 * Semesters - Project Work by Semester
 * Information
 * Experiments - List of speech experiments

Project Notes

 * Unix Notes
 * Active Directory
 * Backups
 * [Network Bridge]
 * Speech S/W Installation
 * Speech Corpus Setup
 * Switchboard Data Notes
 * Experiment Setup
 * Scripts Page
 * Model Building
 * Step 2: Run a Train
 * Step 2: Create the Language Model
 * Step 3: Run the Decode

Network Bridge
Speech recognition tools are awesomely powerful but are also dependent upon software updates and patches. These updates and patches are needed to run the software effectively and properly. Most software relies on an internet connection in order to fetch updates, browse the internet and work remotely. The network team proposed a design in which the UNIX host machine Caesar will be connected to the internet, working as a bridge to allow the 9 Dell PowerEdge servers to have access to the internet.

This poses a problem: Caesar is being used in an environment which only allows one Ethernet connection and is prohibited from using a switch on the UNH Telecom line. The solution that was proposed was to make Caesar work as a bridge and a router. Under this configuration, a switch with a 192.168.X.X environment could be used to provide internet access to the other 10 servers using Sphinx tools. This theory has been proven with a Windows XP networking service. Other problems to consider are; whether OpenSUSE 11.3 (which is Caesars operating system) will take two NIC's (Network Interface Card) and make a bridge, and also if Caesar can act as a router implementing NAT service.

In order for this process to be successful, a wired internet connection is needed. To accomplish this one incoming line from UNH Telecome will connect to Caeser's first NIC and out of Caesar via a second NIC to a switch network environment running 192.168.X.X. Utilizing the Masquerading technique, we should be able to create a network bridge that should implement NAT, but more testing will be needed.

Server
This was successfully implemented on a test system via the following steps:
 * First: Ensure the correct hardware configuration is in place; one machine with two NIC ports, and one machine (or a switch) with one NIC port.
 * Second: Ensure the internet connection on the server machine is working.

1. On the main machine which will serve the internet connection to the rest, open up YAST, select the NIC which will be used to share the connection. NOT THE ONE THAT CONNECTS THE COMPUTER TO THE INTERNET and enter network settings editor.

2. Now modify the settings for this Ethernet device (most likely eth1) to mirror the following settings: (See Figure 1) PAY ATTENTION TO THE SUBNET! The same subnet will have to be used for the client machine.

3. Next, go back to YAST and enter security settings area. Open up firewall settings. Configure the ACTIVE INTERNET connection (probably eth0) as the EXTERNAL FIREWALL ZONE. The "Sharing" interface (probably eth1) should be set to INTERNAL ZONE (See Figure 2).

4. Next, enable IP masquerading, which is in the same area as setting the firewall zones (See Figure 3).

Client

 * On the client machine, the set up is a little bit simpler:
 * 1) Set up a static IP and subnet as done in step 1 for the server. The IP must be different, and the subnet must be the same.
 * 2) Enter the routing tab, and set the IP of the server as the default gateway. This will be the Static IP set up for eth1 on the server. (NOT the IP of the active internet connection on the server).
 * 3) The last step will be to set up the DNS servers to use. The DNS server used by eth0 needs to be used on the server (check the settings of eth0 on the server to get the DNS address).


 * If any more information is needed, refer to this article:
 * http://opensuse.swerdna.org/suseics.html