Disaster may strike anywhere. Transporting data over significant distance beyond the reach of a threatening event preserves data so organizations can rapidly recover in the event of a site going down and Fibre Channel over IP (FCIP), often referred to as “Extension,” is an ideal solution for disaster recovery infrastructure between data centers with fast, continuous, and easy replication of mission-critical data. It was discussed iat length at our FCIA webcast “FCIP (Extension): Data Protection and Business Continuity.” If you missed the live event, you can watch it on-demand at your convenience. We received some great questions at the live event. Here they are with answers from our FCIA experts.
Q. What is the key value proposition of FCIP over other replication technology options?
A. With FCIP, you can go the longest distance with high throughput. With other options, you may have to give a little higher weight to either distance or throughput.
Q. Does it matter if FCIP ISL is running over IP?
A. No, it operates just like a FC ISL. The IP network and associated protocols are invisible to the FC network. The FC network and associated protocol are invisible to the IP network.
Q. What did you mean by frame-based load balancing in FCIP?
A. There are a variety of technologies available for long distance load sharing/balancing with Extension (FCIP):
- Brocade and Cisco both have the equivalent technology to PortChanneling. Brocade calls it “routing”. Neither method is frame-based load balancing. Both companies route or channel traffic across multiple links between the same two domains based on SID/DID (Brocade DBR=Device Based Routing, or Cisco Flow-based) or SID/DID/OxID (Brocade EBR=Exchange Based Routing, or Cisco Exchange Based). Brocade has PBR=Port Based Routing as well. Generally, these methods do not have limitations with varying latency/distance or different bandwidths across the multiple link pathways, additionally, the replication application should be able to successfully recovery from an exchange arriving out of order. Only an exchange could be delivered out of order using these methods. How could an exchange be delivered out of order, you may ask? First, if one of the pathways is significantly longer than the other and the exchange happens to be very short, one exchange may pass another while in transit. Second, one pathway of the IP network may experience congestion requiring some buffering, or packet drops resulting in retransmissions, both cause delivery delays to the ULP (Upper Layer Protocol) and an opportunity for a subsequently transmitted exchange to pass a previously transmitted exchange. Again, it is up to the replication application to adequately deal with this situation, of course, some applications do better than others.
- Brocade offers Extension Trunking (BET), which is a totally different mechanism for transmitting FCIP than the above methods. BET has no limitation on distance/latency. BW is limited to 5x between the min:max on Adaptive Rate Limiting (ARL) on a VE_Port (VE_Port=tunnel), and 4x the lowest to highest maximum values of member circuits belonging to a VE_Port. Nearly all practical environments work within these limits. BET facilitates perfect “Batch” based load balancing. Batches are comprised of multiple FC (16) or FICON (4) frames for greater efficiency during the encapsulation process, referred to as High-Efficiency Encapsulation.
Q. Can you PortChannel or BET to two different Extension boxes?
A. No, a PortChannel or Brocade Extension Trunking (BET) must initiate at a single box and terminate at a single box. Logically, PortChanneling or BET is a point-to-point connection and appears to FSPF as a single ISL with a single route.
Q. Can the two DWDM paths be different in length and still use logical ISL aggregation so that FSPF only sees a single route?
A. Assuming this question is referring to the same subject of the webinar (FCIP), the answer is yes. Multiple IP paths using multiple PortChannel or BET links across any infrastructure including DWDM will appear as a single connection forming a single FSPF route. The links are aggregated using different techniques. PortChanneling and what Brocade calls “DBR”, “EBR” or “PBR” are hash-based algorithms that deterministically sort flows/exchanges onto specific links between two domains. BET performs a batch-based DWRR (Deficit Weighted Round Robin) across the link members of the extension trunk. TCP is used to maintain the order of data sent across all links belonging to that VE_Port.