Luminous E200: Enlightened Transport for Ethernet

Luminous sent me the E200 along with two of its C-Series units so I could create a ringed network. I connected one Gigabit Ethernet transmit-receive pair to each of the C-Series units and connected the C-Series devices to each other. The IP address of each device must be configured via the command line from the RS-232 port on the unit. This is routine, but to access the devices from the LMS, I had tomanually create static routes--full autodiscovery hasn't made its way into the management system yet. I completed the rest of the configuration via a GUI.

As with most carrier-class management systems, Luminous' LMS (Luminous Management System) has a server and a client application. Both are available for Sun Microsystems Solaris 7 or higher and Microsoft Windows NT. I tested under Windows 2000, a platform Luminous hasn't OK'd yet. The company recommends the Solaris version if your network comprises more than a few devices and four or more clients.

LMS supports three types of users: supervisor, administrator and operator. A supervisor can perform all configuration tasks on the units and network. An administrator can only add and delete users. An operator can view configurations but can't change anything.

Let's See Some Traffic

To pass traffic over a packet ring, you must configure paths through the devices. The process is similar to creating a PVC through a carrier's ATM network, but not as easy. Yet LMS is still an early product, so I can overlook many of its idiosyncrasies.

Paths are created by connecting two ports, with each port on a separate Luminous switch. The two ports must be on the same ring, however. If your network has multiple rings, you're in for a workout: You'll need to create a path through each ring. If you have two endpoints on either side of a three-ring network, you'll need one path through each of the three rings. Luminous says the process will be simplified in an upcoming LMS release that will let the system create the entire path across multiple rings "automagically."

Ethernet traffic is divided into three QoS (Quality of Service) categories: best effort, express forwarding or assured forwarding. Best effort is self-explanatory: If it gets there, it gets there. Express gets a higher preference in the queue than assured. Both express and assured forwarding are guaranteed to get to reach their destinations, but express is designed to give time-sensitive data--like voice over IP or packetized video--higher priority.

Rate limiting can also be applied to keep the traffic within the bandwidth defined for the circuit. The three options are off, drop and mark down.

One troubling aspect of LMS is that there is no way to edit an existing circuit. If a user wants to change the bandwidth or QoS, the circuit has to be deleted and re-added. This is cumbersome, and the circuit must be down while changes are being made. Newer LMS versions should correct this.

All this sounds much like other Ethernet switching products until you add the "resilient" part of RPR. Ethernet traffic can be marked as protected or unprotected. Protected circuits are automatically rerouted in the opposite direction the moment the devices realize a ring break has occurred.

I tested this by creating a circuit around the ring with protected Ethernet traffic. I used an Ixia 400 traffic generator with its four-port 10/100 Ethernet blade. With one Ethernet port connected to the E200 and the other connected to one C-Series device, I began blasting traffic. A few seconds later I disconnected the primary pair of Gigabit Ethernet fibers and watched the network go into alarm. By measuring the number of packets received and sent by the Ixia, and the packet size, I was able to deduce that I lost 11 ms's worth of packets on my 10 feet of fiber. In comparison, Sonet systems have a failover rate of less than 50 ms. Luminous cites a sub-50-ms failover for the longest fiber runs.

TDM Traffic

Luminous' implementation of RPR--which it calls Resilient Packet Transport, or RPT--allows the transport of not only Ethernet traffic but TDM (Time-Division Multiplex) traffic as well. So any TDM service (voice, frame relay, private line and so on) can be transported over the ring. The four TDM ports on the E200 can be configured as T1 or E1 circuits. Each port can have different line codings, but they all need to be T1 or E1 ports.

Pools and Policies

As with the Ethernet traffic, the TDM circuits can be set up as protected or unprotected. Using a T1 signal generator, I watched the system fail over the TDM traffic on a protected circuit. One drawback to this system that can't be changed is that during failover the TDM circuit drops and goes into alarm. Once a protected reroute has been completed, the circuit comes back up. This occurred after the 11-ms loss in my tests. If the circuit had been carrying voice traffic, those calls most likely would have been dropped and disconnected. Data traffic should continue unabated once the circuit is back.

SONET Differences

RPR differs from Sonet in the way it deals with the rings. In a typical Sonet network, one ring is used for traffic, the second is protection against failure. This wastes up to 50 percent of the available bandwidth. RPR allows traffic to be placed on either ring. The traffic can be protected or unprotected--it doesn't matter. Either ring will fail over to the other in case of fiber outage. The E220 performs this failover exactly as advertised. This can double the amount of bandwidth available for both transmitting and receiving.

Darrin Woods is a Network Computing contributing editor. He has worked as a WAN engineer for a telecom carrier. Send your comments on this article to him at [email protected].