Brocade Communications Systems StorageWorks 4400 - Enterprise Virtual Array Instrukcja Użytkownika Pobierz pdf (Strona 29)

In previous testing we established that VMFS and Raw Device Mapping perform equally well when

subjected to an Exchange 2007 workload. In this configuration, we configure the Exchange storage with

RDM to leverage the Microsoft Exchange database transportability feature. This Exchange 2007 feature

can easily migrate mailbox databases between servers (physical or virtual) when using a SAN. See the HP

Best practices for deploying an Exchange 2007 environment using VMware ESX 3.5 white paper for

recommendations on configuring Microsoft Exchange 2007 in a VMware ESX environment.

The data in Table 3 compares the differences in I/O performance (response time and I/O request rate)

between the physical server and the VM by using VMFS and RDM storage access management for all of the

logical disks in the two-hour test period.

Table 3. Differences in I/O performance

Database I/O (Vraid 1)

Non-protected VM

Protected VM

(One DR group)

Protected VM

(Two DR groups)

Average database

disk transfers/second

10,454.432

9,671.55

9,752.81

Average database

disk reads/second

5,592

5216

5,352

Average database

disk writes/second

4,862

4454

4,399

Average database

disk read latency (ms)

0.014

Average database

disk write latency (ms)

0.008

0.010

0.009

Log I/O (Vraid 1)

Non-protected VM

Protected VM

(One DR group)

Protected VM

(Two DR groups)

Average database

disk writes/second

2,926

2,426

2,480

Average database

disk write latency (ms)

0.001

The JetStress results indicate that the protected VMs provide comparable storage performance to that of a

non-protected VM when replicating data over a low latency FC inter site link (<1ms). Data replication

affects the application I/O performance in our test configuration by less than 10% due to a slight increase

of the write response times. Using one or two Continuous Access DR group(s) to replicate the five virtual

disks provide similar I/O throughput for the workload tested. However, spreading the EVA virtual disks into

two or more DR groups can optimize full copy or merge operations between the source and destination

arrays.

DR group full copy performance

We measure and compare the DR group normalization (full copy) performance using one and two DR

group(s) to mirror the Exchange Data set. The configuration comprises five virtual disks for a total net

capacity of 2.2 TB. All Vdisks are configured with Vraid 1.

Configuration 1 uses a single DR group to replicate all five Vdisks, Configuration 2 uses two DR groups to

replicate the five Vdisks. The charts in Figure 18 and Figure 19 show the measured throughput during the

normalization window.

1 2 ... 24 25 26 27 28 29 30 31 32 33 34 ... 37 38

Komentarze do niniejszej Instrukcji

Brak uwag

Brocade Communications Systems StorageWorks 4400 - Enterprise Virtual Array Instrukcja Użytkownika Strona 29

Komentarze do niniejszej Instrukcji

Powiązane produkty i podręczniki dla Switche sieciowe Brocade Communications Systems StorageWorks 4400 - Enterprise Virtual Array