Dell AX4-5 Spezifikationen PDF herunterladen (Seite 18)

of auto assign) controls ownership of the LUN in a storage system with two SPs. For more information

about the auto-assign LUN, please see Knowledgebase case emc165941.

VMware native multipathing and failover on ESX 4.0 with CLARiiON

VMware ESX 4.0 contains its own native multipathing software that is built into its kernel. This failover

software, called Native Multipathing Plugin (NMP), has three policies:

• FIXED policy

• Round Robin policy

• Most Recently Used (MRU) policy

On VMware 4.0 servers with CX4 arrays, the FIXED or Round Robin policy is supported.

The FIXED policy on the CX4 provides failback capability. To use the FIXED policy, you must be running

FLARE release 28 version 04.28.000.5.704 or later. Also, the failovermode mode must be set to 4 (ALUA

mode or Asymmetric Active/Active mode). The default failovermode for ESX 4.0 is 1. Use the Failover

Setup Wizard within Navisphere to change the failovermode from 1 to 4.

When using the FIXED policy, the auto-restore or failback capability distributes the LUNs to their

respective storage processors (SPs) after an NDU operation. This prevents the LUNs from all being on a

single storage processor after an NDU. When using the FIXED policy, ensure the preferred path setting is

configured to be on the same storage processor for all ESX hosts accessing a given LUN.

For more details on the benefits of using the Asymmetric Active/Active mode with CLARiiON storage

systems, please see EMC CLARiiON Asymmetric Active/Active Feature (ALUA) available on Powerlink.

With the FIXED policy, there is some initial setup that is required to select the preferred path; the preferred

path should also be the optimal path when using the ALUA mode. If set up properly, there should not be

any performance impact when using failovermode 4 (ALUA). Note that FIXED sends I/O down only a

single path. However, if you have multiple LUNs in your environment, you could very well choose a

preferred path for a given LUN that is different for other LUNs and achieve static I/O load balancing.

FIXED performs an automatic restore, hence LUNs won't end up on a single SP after an NDU.

When using Round Robin there is no auto-restore functionality, hence after an NDU all LUNs will end up

on a single SP. A user would need to manually trespass some LUNs to the other SP in order to balance the

load. The benefit of Round Robin is that not too many manual setups are necessary during intial setup; by

default it uses the optimal path and does primitive load balancing (however, it still sends I/O down only a

single path at a time). If multiple LUNs are used in the environment, you might see some performance

boost. If you had a script that takes care of the manual trespass issue, then Round Robin would be the way

to avoid manual configuration.

On a CX3 or earlier CLARiiON storage systems, the Most Recently Used (MRU) or Round

Robin policy must be used with failovermode=1.

Note that the Most Recently Used (MRU) and Round Robin policies do not provide failback capability.

Furthermore, the Round Robin policy does not provide true dynamic load balancing; it sends I/O down one

chosen path at a time alternating through paths on the owning SP. The path selected for I/O is controlled by

the Round Robin algorithm. The FIXED and MRU policies also send I/O down only a single selected path

for a given LUN, unless that path becomes unavailable.

Figure 9 shows how the FIXED policy is configured with the CX4 storage system using VMware’s NMP

software.

EMC CLARiiON Integration with

VMware ESX Server

Applied Technology 18

1 2 ... 13 14 15 16 17 18 19 20 21 22 23 ... 57 58

Kommentare zu diesen Handbüchern

Keine Kommentare

Dell AX4-5 Spezifikationen Seite 18

Kommentare zu diesen Handbüchern

Verwandte Produkte und Handbücher für Systemverwaltungssoftware Dell AX4-5