The figure below shows a system configuration example of a standby-less system switchover (effects distributed) system. When an error occurs in the regular unit, processing intended for the back-end servers in the primary system where the error occurred is divided in units of back-end servers and assumed by multiple running server machines, where it is executed.

Figure 26-80 System configuration example of a standby-less system switchover (effects distributed) system

Explanation

1. When an error occurs in unit 1, BES1 executes its processing in unit 2 as a guest BES, and BES2 executes its processing in unit 3 as a guest BES.
2. If an error occurs in unit 2 while unit 1 is still in error status, BES1, BES2, BES3, and BES4 all perform their processing in unit 3 as guest BESs.

Although the maximum number of units that can be defined in an HA group is 32, by defining multiple HA groups, you can use the standby-less system switchover (effects distributed) facility in a large system with more than 32 units. The following figure shows a system configuration example of a standby-less system switchover (effects distributed) system when multiple HA groups are defined.

Figure 26-81 System configuration example of a standby-less system switchover (effects distributed) system when multiple HA groups are defined

Explanation

In HA group 1, 32 units (unt1 through unt32) are defined. If an error occurs in the server machine of unt1, BES1 executes processing as a guest BES of a unit within HA group 1. The same holds true for HA groups 2 and 3.

If three HA groups are defined, the standby-less system switchover (effects distributed) facility can be used in a 96-unit system.

When you use the standby-less system switchover (effects distributed) facility, HiRDB takes the following issues into consideration in determining the switching destinations for the individual servers:

• Balancing among the units the number of running guest BESs
• Balancing among the units the number of servers that share each global buffer

When the standby-less system switchover (effects distributed) facility is used, HiRDB uses the following procedure to determine switching destinations.

Procedure

1. Determine the switching-destination unit for the server that has the highest priority among all servers. The following are the determination criteria:
   Balancing of the number of servers to be switched from one unit to other units
   Balancing among the units at the switching destination the number of servers that will share a global buffer
   
2. Repeat steps 3 through 9 from i = 1 to [number of units in the HA group - 2].
3. Select a server with a priority ranking of i + 1 whose switching destination has not been determined.
4. If there is an applicable server, proceed to step 5. If there is no applicable server, set i equal to i + 1 and return to step 3.
5. Assume that switching destinations 1 through i must be determined for the selected servers, and that the unit to which these servers belong is the unit where the error occurred.
6. Select units other than the unit where the error occurred as switching destination candidates.
7. From the servers belonging to the unit where the error occurred, select all servers that have switching destination candidates as defined switching destinations. Assume that guest BESs for the server in question are allocated to the switching-destination unit with the highest priority.
8. From the servers belonging to the unit where the error occurred, select all servers that do not have switching destination candidates as defined switching destinations, and determine the i + 1^th switching-destination unit for each server. The following are the determination criteria:
   The unit with the smallest number of guest BESs
   Units in which fewer guest BESs share a global buffer with the server
   
9. Return to step 3.

(a) Configuration with four units

The standby-less system switchover (effects distributed) facility is applied to the following configuration consisting of four units:

In this case, a global buffer is shared in each of the following groups:

• Group A: BES1A, BES2A, BES3A, and BES4A
• Group B: BES1B, BES2B, BES3B, and BES4B
• Group C: BES1C, BES2C, BES3C, and BES4C

1. Determining the switching-destination unit with the highest priority
   An example is shown below that determines the switching-destination unit with the highest priority for BES3B. A switching destination is determined so that it does not coincide with servers that share the same global buffer (BES1B and BES2B) or BES3A. Specifically, unt1 is selected because the switching destination of BES1B (unt3), the switching destination of BES2B (unt4), and the switching destination of BES3A (unt4) cannot be selected.
   

   Server	Host BES	Highest priority	Second priority	Third priority
   BES1A	unt1	unt2	Allocated	Allocated
   BES1B		unt3	Allocated	Allocated
   BES1C		unt4	Allocated	Allocated
   BES2A	unt2	unt3	Allocated	Allocated
   BES2B		unt4	Allocated	Allocated
   BES2C		unt1	Allocated	Allocated
   BES3A	unt3	unt4	Allocated	Allocated
   BES3B		unt1	Allocated	Allocated
   BES3C		Allocated	Allocated	Allocated
   BES4A	unt4	Allocated	Allocated	Allocated
   BES4B		Allocated	Allocated	Allocated
   BES4C		Allocated	Allocated	Allocated

2. Determining the switching-destination unit with the second or subsequent priority
   An example is shown below that determines the switching-destination unit with the second priority for BES2B.
   First, unt4, which is a defined switching destination of BES2B, and unt2, to which BES2C belongs, are assumed to be units where the errors occurred (step 5), and unt1 and unt3 are selected as switching-destination candidates (step 6). Next, from the servers belonging to the units where the errors occurred (unt2 and unt4), servers (BES2A, BES2C, BES4A, and BES4C) that have the switching destination candidates as defined switching destinations are extracted, and it is assumed that guest servers are allocated as shown below (step 7).
   BES2A: unt3; BES2C: unt1; BES4A: unt1; and BES4C: unt3
   Next, servers (BES2B and BES4B) that do not contain the switching destination candidates as defined switching destinations are extracted, and the switching-destination unit with the second priority is determined. In this case, the number of guest BESs is the same for unt1 and unt3, and BES2B and BES4B share a global buffer. Therefore, BES2B is switched to unt1 and BES4B is switched to unt3.
   

   Server	Host BES	Highest priority	Second priority	Third priority
   BES1A	unt1 Switching-destination candidate unit	unt2	unt4	Allocated
   BES1B		unt3	unt2	Allocated
   BES1C		unt4	unt2	Allocated
   BES2A	unt2 Unit where an error occurred	unt3	unt1	Allocated
   BES2B		unt4	unt1	Allocated
   BES2C		unt1	Allocated	Allocated
   BES3A	unt3 Switching-destination candidate unit	unt4	Allocated	Allocated
   BES3B		unt1	Allocated	Allocated
   BES3C		unt2	Allocated	Allocated
   BES4A	unt4 Unit where an error occurred	unt1	Allocated	Allocated
   BES4B		unt2	Allocated	Allocated
   BES4C		unt3	Allocated	Allocated

The following table provides an example of a switching destination definition in a 4-unit configuration.

Table 26-20 Switching destination definition example in a 4-unit configuration

Server	Host BES	Highest priority	Second priority	Third priority
BES1A	unt1	unt2	unt4	unt3
BES1B		unt3	unt2	unt4
BES1C		unt4	unt2	unt3
BES2A	unt2	unt3	unt1	unt4
BES2B		unt4	unt1	unt3
BES2C		unt1	unt3	unt4
BES3A	unt3	unt4	unt2	unt1
BES3B		unt1	unt4	unt2
BES3C		unt2	unt4	unt1
BES4A	unt4	unt1	unt3	unt2
BES4B		unt2	unt3	unt1
BES4C		unt3	unt1	unt2

(b) Configuration with five units

The standby-less system switchover (effects distributed) facility is applied to the following configuration consisting of five units:

In this case, a global buffer is shared in each of the following groups:

• Group A: BES1A, BES2A, BES3A, BES4A, and BES5A
• Group B: BES1B, BES2B, BES3B, BES4B, and BES5B
• Group C: BES1C, BES2C, BES3C, BES4C, and BES5C
• Group D: BES1D, BES2D, BES3D, BES4D, and BES5D

The following table provides an example of a switching destination definition in a 5-unit configuration.

Table 26-21 Switching destination definition example in a 5-unit configuration

Server	Host BES	Highest priority	Second priority	Third priority	Fourth priority
BES1A	unt1	unt2	unt5	unt3	unt4
BES1B		unt3	unt4	unt2	unt5
BES1C		unt4	unt3	unt5	unt2
BES1D		unt5	unt2	unt4	unt3
BES2A	unt2	unt3	unt1	unt4	unt5
BES2B		unt4	unt5	unt1	unt3
BES2C		unt5	unt4	unt3	unt1
BES2D		unt1	unt3	unt5	unt4
BES3A	unt3	unt4	unt2	unt1	unt5
BES3B		unt5	unt1	unt2	unt4
BES3C		unt1	unt5	unt4	unt2
BES3D		unt2	unt4	unt5	unt1
BES4A	unt4	unt5	unt3	unt1	unt2
BES4B		unt1	unt2	unt3	unt5
BES4C		unt2	unt1	unt5	unt3
BES4D		unt3	unt5	unt2	unt1
BES5A	unt5	unt1	unt4	unt2	unt3
BES5B		unt2	unt3	unt1	unt4
BES5C		unt3	unt2	unt4	unt1
BES5D		unt4	unt1	unt3	unt2