STORAGE TECHNOLOGYThe extended link serviceSequence RetransmissionRequest is sent by the initiator(the host) to the target deviceto request that a targetretransmit data back to theinitiator, or to inform thetarget that data will beretransmitted back to thetarget. The invocation andacceptance of an SRR indicatethat a previous commandor data frame was lost.target during a read operation.The DTC is importantin determining whether alldata was successfully transmittedor received. TheE_STAT word value has twobit values that identify thestate of the exchange. Thefirst value is the Completionbit, which specifies whetherthe exchange is still open;and the second value isthe Sequence Initiative bit,which indicates which portholds the initiative (that is,which end of the transmissionis supposed to make thenext move).Sequence RetransmissionRequest functionThe next part of FC-TAPEerror recovery involves theextended link service SRR.This service is sent by the initiator(the host) to the targetdevice to request that a target retransmit data back to the initiator,or to inform the target that data will be retransmitted back to thetarget. The invocation and acceptance of an SRR indicate that aprevious command or data frame was lost.The SRR payload consists of an OX_ID, a Relative Offsetvalue, and a Routing Control (R_CTL) value, all of which assistin determining which data needs to be retransmitted. The RelativeOffset field identifies an offset value that the initiator requires fordata retransmission, and the R_CTL field defines the frame functiontype that is to be retransmitted. The R_CTL value identifieswhether the retransmission is a data descriptor (FCP_XFER_RDY),command status (FCP_RSP), or solicited data (FCP_DATA). If thetarget accepts the SRR, the target will respond with an ACC, whichwill then allow the data to be retransmitted.Confirmed Completion (FCP_CONF) is used to acknowledgesuccessful receipt of a response frame. In other words, once thetarget sends an FCP_RSP to the initiator, the initiator will send backconfirmation informing the target that the response was successfullyreceived. The confirmation is merely an added protection mechanismto help ensure successful delivery of the response—allowingthe target to safely close the exchange. Confirmed Completion ina Fibre Channel environment is equivalent to the SCSI functionREQ/ACK for command completion.Example error recovery implementation with FC-TAPETo demonstrate the error recovery process related to FC-TAPE,the Figure 2 example shows how a data frame (FCP_DATA), lostduring a SCSI write command (FCP_CMD), is recovered usingFC-TAPE as follows:• Section A: An uninterrupted flow process writes data to atape device. Once the data is successfully transmitted to thetape device, the tape device returns good status back to theinitiator.• Section B: An error causes the third data frame to be lost.In this instance, the initiator is waiting for a response, andthe target is still waiting for more data. During this deadlockperiod, the three-second REC_TOV timer expires. Sensing adelay in the response from the target, the initiator then transmitsa REC requesting information from the target regardingthe status of the exchange. The target responds to the RECby transmitting an ACC back to the initiator. The ACC payloadfrom the target contains the DTC value, indicating thatthe number of bytes successfully received by the target wasincomplete. Next, the E_STAT bit in the ACC payload informsthe initiator that the exchange is still open, and that thetarget does not own the sequence initiative. In other words,the target is telling the initiator to take the next action to continuethe exchange.• Section C: To correct the error condition, the initiator willtransmit an SRR to the target requesting that the targetresend an FCP_XFER_RDY (request to allow data transfer).The specific request for the FCP_XFER_RDY is defined inSection A Section B Section CInitiatorTarget InitiatorTarget InitiatorFCP_CMD (write)FCP_CMD (write)SRRFCP_XFER_RDYFCP_DATAFCP_RSPFCP_XFER_RDYFCP_CMD (write)012 xxxxxxxxxxxxxxREC (REC_TOV)ACCFigure 2. Example FC-TAPE error recovery process012ACCFCP_XFER_RDYFCP_DATAFCP_RSPTarget90DELL <strong>POWER</strong> <strong>SOLUTIONS</strong> Reprinted from Dell Power Solutions, August 2005. Copyright © 2005 Dell Inc. All rights reserved. August 2005
STORAGE TECHNOLOGY• Line 4: The initiator issues an SRR.• Line 5: The target replies toFigure 3. Example Fibre Channel trace showing FC-TAPE error recoverythe SRR.• Line 6: The target resends the“lost” FCP_RSP (good status).• Line 7: The initiator sends anFCP_CONF to the target indicatingthat it received the FCP_RSP.the R_CTL payload in the SRR. The target will respond tothe SRR with an ACC and then transmit an FCP_XFER_RDYback to the initiator, indicating that the target is ready toreceive data transmission. Once the initiator receives theFCP_XFER_RDY, the initiator will immediately retransmitthe FCP_DATA data frames.The amount of time to recover from the original lost data frameis no more than three seconds (the REC_TOV), which is typically farless time than the error recovery mechanisms used by tape backupsoftware. For this reason, FC-TAPE can be a valuable asset in linklevelerror recovery procedures for tape backup devices.Fibre Channel trace showing FC-TAPE error recoveryFigure 3 displays a segment from a Fibre Channel trace in whichFC-TAPE error recovery is invoked because of a lost FCP_RSP(good status). The debug trace data shown in Figure 3 correspondsto the following process:• Line 1: The host tape backup server issues a SCSI readcommand.• Line 2: No FCP_RSP is received within the REC_TOV timer(three seconds), and thus a REC is transmitted. (The SCSIread command was issued at the 3.4 second mark, and theREC was issued at the 6.3 second mark.)• Line 3: The target accepts the REC by responding with an ACC.This example demonstrates the efficiency of FC-TAPE in performingerror recovery without incurring overhead on the tapebackup software. FC-TAPE error recovery is conducted at the FibreChannel link level and completed within seconds between the initiatorand the target tape device.Recovering from tape backup errors quickly and efficientlyBest practices for tape backup operations recommend that administratorsenable FC-TAPE error recovery on the Fibre Channel adapteras well as the data verification setting in the tape backup software.By combining FC-TAPE and tape backup software error recovery,administrators can quickly and easily enhance the capability todetect and correct errors on a Fibre Channel link, thereby helpingto preserve data integrity.Richard Golasky is a development engineer senior consultant in theDell/EMC Enterprise Storage Group. His responsibilities include all areasof tape backup, including Fibre Channel storage, high-availability clusters,network attached storage, and performance analysis. He has a B.S. inElectrical Engineering from Florida Atlantic University.FOR MORE INFORMATIONDell tape backup devices:www.dell.com/tapebackupExperienceDell Power Solutionswww.dell.com/powersolutions Reprinted from Dell Power Solutions, August 2005. Copyright © 2005 Dell Inc. All rights reserved. DELL <strong>POWER</strong> <strong>SOLUTIONS</strong> 91