bymdetrick05-06-201508:00 AM - edited 05-06-201511:55 AM
IPEX is based on 5 pillars. These tenants are distinctly different, crucial to many environments, and made up the design criteria for Brocade Extension Engineering. To review, they are:
High Performance (covered in part 2 of 3 in this series)
Today, we will discuss the remaining four pillars.
You can think of your IPEX network as a logical IP storage network that underlies the physical IP network and WAN. Monitoring, alerting and diagnostic tools are part of that logical network. Brocade has various products that provide in-depth visualization and management of the IP storage network. For instance, MAPS (Monitoring Alerting Policy Suite), which is a preconfigured monitoring and alerting suite with preconfigured groups, preconfigured thresholds and preconfigured actions. You just turn it on. MAPS is part of Fabric Vision, which is a synergistic system between FabricOS, the switching ASIC and the management platform (Connectrix Manager Converged Network Edition). The base Brocade 7840 Extension Switch includes the Enterprise License, so, there are no additional licenses to purchase for Fabric Vision.
FC Throughput (Rx & Tx)
Round Trip Time (RTT)
The extension dashboard monitors every aspect of the IP network that we can think of. If the IP network is not performing to the agreed upon SLA with your Network Administrators, you will know immediately. Upon being alerted to a possible issue in the IP network, you have Wtool (WAN Test Tool) available to test your circuits. Wtool will generate test patterns across a circuit in a non-disruptive manner. Best practice; design your Brocade extension network with more than one circuit. For example, one circuit to data center LAN switch 1 and a different circuit to data center LAN switch 2. These two circuits are members of the same trunk. This way when the data center Network Admins take down a LAN switch you don’t lose connectivity. When Wtool tests a circuit, the tunnel remains operational just minus the bandwidth of the circuit in test, plus no data is lost or delivered out of order to the ULP. Merely, a circuit is put into test and upon completion put back into service after which another circuit can be tested. A report is generated and the results of the previous run are available for comparison.
The same availability Brocade builds into FCIP, which is lossless and guarantees in-order deliver, has been applied to IPEX. This is called Extension Trunking and it’s not new. Brocade has been doing this for years and we know it works great. As long as one path remains online, the trunk will remain operational and ARL may adjust the rate limiting to maintain bandwidth. Circuits that terminate at a single VE_Port are always lossless and have in-order delivery. This includes backup circuits that are designated “Metric 1”. Metric 0 are the normally active circuits. If all Metric 0 within a group go offline, the metric 1 circuit becomes active. Failover and failback are automatic.
This is a typical architecture using Extension Trunking. For more information, please refer to the Technical Brief, “Extension Trunking” at
IPEX customers have deployed for no other reason than to gain high-speed encrypters for their IP storage environment. More often than not, the IP storage end-devices either cannot perform IPsec or have a prohibitive penalty when performing IPsec. The 7840 IPEX offloads this burden from the IP storage end-devices.
Brocade 7840 IPsec has a hardware implementation. Being implemented in HW allows line rate operation and very little added latency (a couple of µs). Brocade offers IPsec as part of the 7840 base unit, there is no additional cost or licenses.
Customers told Brocade during the design phases of IPEX that if they would be required to change cabling, IP addresses, subnets, or VLANs, that would result in a “non-starter”. We had no reason to doubt this. IPEX requires no change in any of these things.
The way LAN side deployment of IPEX works is as such… The Brocade 7840 is Link Aggregated or LAG’ed (802.1AX) to one or more existing data center LAN switch. We do LAG because it provides multiple redundant connections with lots of bandwidth. These are 10GE connections. The connections use 802.1Q VLAN Tagging, therefore, they can carry the traffic from multiple VLANs. On the Brocade 7840 side we have a SVI (Software Virtual Interface) that is seen behind the LAG connection. On the IP storage end-devices, changing the gateway addresses to the SVI address for those routes associated with just the subnets on the remote side, directs IP storage traffic to the Brocade IPEX managed tunnel. We cannot think of any way to make this simpler.
The WAN facing side of the IPEX tunnel remains the same as an FCIP tunnel.
Thank you so much for reading the entire series (Part 1 and Part 2). May the remainder of EMC World be great for you and safe travels.