Transcript
Application Note
Using the Falcon Dual Controller EonStor A16F-R1211 Hub Embedded Storage Subsystem
Date:
June 8, 2004
Application Note
Using the Hub Embedded A16F-R1211 /
This page is left intentionally blank.
Application Note
Using the Hub Embedded A16F-R1211 /
Using the Raid Inc Hub Embedded Fibre to SATA unit The Raid Inc Fibre to SATA subsystem with the embedded hub can be used in a variety of configurations. In one particular configuration a fully redundant system can be designed, which provides resilience all the way from the Host Bus to Adaptec, to the host cables, through the SFP connectors, right down to the controller. Another configuration allows for capacity scaling providing hundreds of Terabytes to users. There are two modes of operation – Hub enabled and Hub disabled. The first section of this document discussed Hun enabled configurations.
Hub enabled Configurations Ensure that the hub is enabled by setting switch 4 on the rear DIP switch to the down position as shown in Figure 1 below. Figure 1 Enabling the embedded hub
There are 5 ports per controller (4 host, 1 expansion). An expanded view of the ports is shown below in Figure 2. Figure 2 Port view
Out Chan0 Chan1 EXP Chan0 Chan1 In
Prim a ry Controller
Sec onda ry Controller
Page 3 of 9
Application Note
Using the Hub Embedded A16F-R1211 /
Each set1 of controller ports are connected together via Port Bypass Circuitry, so when both controllers are inserted all four channel 0 ports are connected together and all four channel 1 ports are connected together.
Mode 1 – Dual Server configuration In the diagram shown in
Figure 3 two logical drives are mapped to each controller. Logical drive 0 has been mapped to the primary controller and logical drive 1 has been mapped to the secondary controller through the configuration utility. Two servers with single Fibre Channel Host Bus Adapters are connected to the RAID sub-system, the first server is connected to the channel 0 (in) port of the primary controller and the second server is connected to the channel 1 (in) port of the primary controller. In this configuration each server will only see one logical drive. The server on the left can access logical drive 0 and the server on the right will have access to logical drive 1. This type of configuration could be within a heterogeneous environment where one server runs a UNIX system and the other could be Microsoft based If the primary controller fails then the system will continue to deliver uninterrupted I/Os since logical drive 0 has now failed over to the secondary controller and because of the hub functionality the server is connected to channel 0 of the secondary controller. If the secondary controller fails then in a similar manner the system will continue to deliver uninterrupted I/Os since logical drive 1 has now failed over to the primary controller. Note that even though the connections are physically shown on the primary controller this is arbitrary as the primary controllers connections are connected to the secondary controllers connections. It is important to recognize any limitations with the configurations. In this mode if the primary controller should fail then when it is eventually replaced, there will obviously be a break in the data path, so downtime will need to be scheduled at some stage. The important point however, is that the downtime can be planned rather than occurring unexpectedly. The Host/LUN to Logical Drive Mapping is shown in
Table 1 for each of the configurations.
1
All channel 0 ports are connected together and all channel 1 ports are connected together.
Page 4 of 9
Application Note
Using the Hub Embedded A16F-R1211 /
Figure 3 Dual server
Prim a ry Controller
Logical Drive 0 Mapped to Primary controller
Sec onda ry Controller
Logical Drive 1 Mapped to sec ondary controller
Mode 2 – Single Server/ dual HBA A single server/dual HBA configuration is shown in Figure 4. In this example the system will have access to both logical drives. Drive mapping is set up so that logical drive 0 is mapped to the primary controller and logical drive 1 is mapped to the secondary
Page 5 of 9
Application Note
Using the Hub Embedded A16F-R1211 /
controller. If a controller should fail then the drives can be accessed through the surviving controller. This system could use the two HBA’s to enhance performance. Figure 4 Single server/dual HBA configuration
Prim ary Controller
Logical Drive 0 Mapped to Prima ry c ontroller
Sec ondary Controller
Logical Drive 1 Mapped to sec ond a ry c ontro ller
Page 6 of 9
Application Note
Using the Hub Embedded A16F-R1211 /
Mode 3 This system shown in Figure 4 is resilient to controller failures, however in this configuration, HBA, SFP or cables failures are not taken into account. This can be achieved however by implementing the configuration shown in Figure 5. The server has two dual ported Host Bus Adapters directly connected to the input ports of the controllers. This mode of operation requires Multi-Pathing software such as Veritas Volume Manager since otherwise the server would “see” four logical drives under normal conditions. Figure 5 Implementing a fully redundant configuration.
Logical Drive 0 Mappedto Primary controller
Prim ary Controller
Sec ondary Controller
Logical Drive 1 Mapped to sec ondary controller
The host LUN Mapping data for each mode is shown in the table below: Page 7 of 9
Application Note
Using the Hub Embedded A16F-R1211 /
Table 1 Host/LUN Mappings
Mode 1 Mode 2 Mode 3
Primary Controller Ch0 ID 102 (LD0) ID 102 (LD0) ID 102
Primary Controller Ch1 NA NA ID 104
Secondary Controller Ch0 NA NA ID 103
Secondary Controller Ch1 ID 103 (LD1) ID 103 (LD1) ID 105
Hub Disabled configurations It is also possible to scale in terms of capacity. In this situation, subsystems are daisy chained going from the out ports of one system to the in ports of the next system. In the mode the hub must be disabled and by setting switch 4 (refer to Figure 1) to the up position. JBOD systems can also be added to each RAID enclosure by connecting to the expansion port (refer to Figure 2) of the RAID controllers. In the example shown in Figure 6 there are two servers connected to the in ports of one subsystem (subsystem1). Subsystem 1 is connected to subsystem2 through subsystem1’s out ports. Both subsystem1 and subsystem2 have a single JBOD system attached. Further subsystems and JBODs can be attached giving extremely high capacity configurations.
Page 8 of 9
Application Note
Using the Hub Embedded A16F-R1211 /
Figure 6 Disabling the hub to provide capacity scaling.
Page 9 of 9