Brocade Communications Systems StorageWorks 4400 - Enterprise Virtual Array Instrukcja Użytkownika Pobierz pdf (Strona 16)

VMs in serial order because these VMs generally have dependencies between each other, and because we

want the most predictability. The medium and low priority sections start their VMs in parallel to provide the

best RTO. VMs will be brought up in parallel based on the number of ESX hosts that you have. If there are

four ESX hosts at the recovery site then four medium and low priority VMs can come up at once. If there is

only one ESX host at the recovery site then virtual machines will be brought up one VM at a time. The

number of parallel startups is limited to 16 even when there are more than 16 ESX hosts.

Configuring data replication

Understanding how Site Recovery Manager integrates with array-based replication components is key to

successful deployment.

With Site Recovery Manager, the smallest possible unit of storage for data replication is a LUN. LUNs are

EVA Vdisks presented to the ESX hosts. Vdisks are the unit of replication for the arrays; however,

Continuous Access EVA uses a concept of consistency groups (DR group) to replicate Vdisks. Therefore, a

DR group represents the smallest possible granularity for failover.

It is not possible to fail over a single Vdisk/LUN in a multi member DR group without failing over all the

group members. This means that you must group virtual machines on LUNs, and you must also plan the

grouping of the Vdisks into DR groups. A DR group can contain virtual disks from multiple EVA disk groups,

but all DR group member virtual disks must be in the same array and use the same preferred controller on

the array.

The EVA provides a limited number of DR groups

so understanding your environment and its replication

granularity requirements can help you reduce the number of DR groups required for your environment and

provide improved efficiency.

VMware formats LUNs with VMFS to store virtual machines. These VMFS -formatted LUNs are referred to as

datastores. Datastores commonly contain only a single LUN but do have the ability to span across multiple

LUNs. As an alternative, administrators can use raw device mapping to attach a LUN directly to the virtual

machine. In this case, the LUN is formatted with the native file system of the VM. With Site Recovery

Manager update 1, VMFS and Raw Device Mapping (virtual and physical) are supported for data

replication.

The smallest group of datastores (and therefore LUNs) that can have its contents failed over with Site

Recovery Manager is referred to as a datastore group. These groupings are calculated for you so that you

do not have to worry about figuring them out. In Figure 9, LUN 1 is formatted with VMFS, and the volume is

named VMFS1 and is shared by three virtual machines. It has no other dependencies and sits in its own

datastore group.

Note: Max number of DR groups may vary, depending on the EVA firmware.

Consult the Continuous Access EVA release notes for the number of DR groups.

1 2 ... 11 12 13 14 15 16 17 18 19 20 21 ... 37 38

Brak uwag

Brocade Communications Systems StorageWorks 4400 - Enterprise Virtual Array Instrukcja Użytkownika Strona 16