Dell PowerVault DL2100 User Manual

Dell
PowerVault
DL2000
Backup
Performance

A Dell Technical White Paper

Dell PowerVault DL2000 Powered By
Symantec

Muffadal Quettawala and
Scott Reichmanis

Introduction

The Dell PowerVault ™ DL2000 – Powered by Symantec Backup Exec

is a exible, cost-effective backup solution for simple, manageable

data protection.

This whitepaper evaluates the throughput performance capabilities

of the PowerVault DL2000 in its two hardware congurations – Value

Series and Performance Optimized1. With this information, the

end-user will be able to accurately size their backup infrastructure

to meet backup window requirements by scaling the number of

DL2000s needed for their environment and by optimizing backup

schedules and balancing backup workloads.

The PowerVault DL2000 Performance Optimized version was

evaluated in a six (6) Gigabit Ethernet (GbE) port conguration and

the DL2000 Value Series version was evaluated in both a six (6) GbE

port and a two (2) GbE port conguration. The backup client system

infrastructure consisted of a Microsoft® Exchange Server 2007

environment and a Microsoft® SQL Server® 2008 environment as

illustrated in Figure 1 below.

Figure 1 - Backup Infrastructure Environment

1

The DL2000 – Powered by Symantec also comes in four (4) different software editions but they have no

impact on backup throughput, and hence are not compared in this whitepaper.

The maximum backup throughput achieved for each of the three

system congurations is summarized in the table below:

Conguration End-to-end Effective

Backup Throughput for
Microsoft Exchange 2007
(GB/hr)

PowerVault DL2000
Performance

Optimized with six (6)
GbE ports

PowerVault DL2000
Value Series with six

(6) GbE ports

PowerVault DL2000
Value Series with two

(2) GbE ports

1833 2645

1578 2218

991 1403

End-to-end Effective
Backup Throughput

for Microsoft SQL

2008 (GB/hr)

These results provide an end-user with valuable data and insight for

planning, sizing and using the DL2000 Backup to Disk Appliance

within their IT infrastructure.

Test Setup

The test setup details are summarized in the table below:

Backup

Solution

DL2000 – Powered
by Symantec
Backup Exec with
Performance
Optimized hardware

and 6 GbE ports

DL2000 – Powered
by Symantec Backup
Exec with Value
Series hardware and

6 GbE ports

DL2000 – Powered
by Symantec Backup
Exec with Value
Series hardware and

2 GbE ports

1

1

1

No. of

MD1000

Enclosures

1

2

5

1

2

5

1

2

2

Standard

Standard

Standard

Backup Type Clients

6, 12 and
24 Microsoft
Exchange

Software

Compression

Software

Compression

Software

Compression

Software

Encryption

Software

Encryption

Software

Encryption

2007 clients
with average

Information
Store size of
approximately

36 GB

6, 12 and 24

Microsoft SQL

2008 clients
with average

database
size of
approximately

36 GB

Note: The 6, 12
and 24 clients
were used
for 1, 2, and
5 enclosure
benchmarks
respectively.

3,4,5

5

1

The DL2000 NIC ports were set up in a Smart Load Balanced (SLB) team using the Broadcom Advanced Control Suite

(BACS) with receive and transmit buffer sizes set to 1024. TCP Ofoad capabilities are not available to a team with
heterogeneous network adapters. To enable TOE, congure the DL2000 with each GbE port assigned to a separate

subnet. The Dell PowerConnect 6248 10/100/1000BASE-T Ethernet switch was used to connect the DL2000 with the
client systems in a dedicated network environment.

2

Each MD1000 enclosure contained 15 1TB 7.2K SATA disk drives and disk groups were automatically congured via

the Symantec Backup Exec Storage Provisioning Option (SPO).

3

The clients ran on Windows Server 2008 64-bit Standard Edition. Synthetic database load for the backup system

clients was generated using the Microsoft Exchange Load Generation tool, version 08.02.0045.000 for Exchange
clients and the RedGate SQL Data Generator version 1.2 for SQL clients.

4

The size of the total backup set was sufcient to eliminate in-memory caching effects and maintain sustained

performance through a 10 minute measurement interval

5

Backups were performed in a homogenous application server environment, i.e. the environment consisted of only

Exchange Servers or only SQL Servers). The Exchange Information Stores and SQL databases were backed up in a static

environment, i.e. no transactional database activity was taking place during backup operations.

For the DL2000 - Powered by Symantec Backup Exec, the following

additional conguration settings were used when performing

backups:

• Each backup client system was congured as a separate backup

job

• The maximum number of concurrent jobs per virtual disk was set

to one (1).

• Consistency checks and backup verication were disabled on all

backup jobs to minimize pre- and post-processing backup times.

• For Software Encryption backups, 128-bit AES keys were used.

• For Software Compression backups, Backup Exec’s software

compression feature was used. Note: For SQL 2008, the SQL

Server 2008 Enterprise Edition software compression feature

was disabled.

The next section provides a detailed summary of the throughput

results obtained for Microsoft Exchange 2007 and Microsoft SQL

2008 clients respectively.

Results and Analysis

The gures below compare the end-to-end backup throughput results

obtained for PowerVault DL2000 - Powered by Symantec Backup

Exec with Microsoft Exchange and Microsoft SQL clients. The results

compare systems with one (1), two (2) and ve (5) enclosures for

three (3) different backup types - Standard, Software Compression

and Software Encryption.

Microsoft Exchange 2007 Microsoft SQL 2008

The maximum backup throughput achieved for each of the three

system congurations is summarized as follows:

Backup

Solution

DL2000
Performance

Optimized with 6
GbE ports

DL2000 Value

Series with 6
GbE ports

DL2000 Value
Series with 2

GbE ports

1

Adding clients beyond 12 reduced throughput due to processor limitations on the DL2000 Value Series.

2

For DL2000 Value Series with two (2) GbE ports, throughput was limited by the two (2) GbE ports. Increasing

the number of MD1000 enclosures and backup client systems had little to no effect on the backup throughput.

3

Throughput was measured at each of the NIC ports of the DL2000 during a 10 minute interval of ingest.
Values were adjusted by 4.3% to account for the Ethernet frame preamble, Ethernet header and TCP/IP
Packet header overhead. Compression ratios were factored, where applicable in the aggregate throughput
measurements to arrive at end-to-end effective throughput. Software compression ratios were 1.2 for

Exchange and 1.7 for SQL.

2

Number of

Backup client

systems

24 5 1833 2645

1

12

12 2 991 1403

Number of

MD1000

Enclosures

2 1577 2218

Exchange

Throughput3

GB/hr

SQL

Throughput3

GB/hr

As seen in the table above, backup throughput rates for the DL2000

Performance Optimized version with six (6) GbE ports were found to

be up to 17% faster than DL2000 Value Series with six (6) GbE ports

and up to 47% faster than DL2000 Value Series with two (2) GbE

ports.

The following discussion analyzes the data gathered during the

benchmark from three different perspectives.

Analysis by the Number of MD1000 Enclosures:

In general, going from one to two enclosures provided a signicant

increase in throughput. Beyond two enclosures the benets were

minimal. With the Value Series hardware, going beyond two (2)

MD1000 enclosures resulted in a small decrease in throughput in a

few cases since the processor became the limiting factor.

Analysis by number of GbE ports:

For the six (6) GbE port systems the backup throughput was never

network limited. For the PowerVault DL2000 Value Series hardware

with only two (2) GbE ports, backup throughput rates were network

limited in all cases where there were two (2) or more enclosures. For

the one (1) enclosure case, there were not enough clients to saturate

the NIC bandwidth

Analysis by Backup Type:

Software compression and software encryption are performed

by the backup clients and thus do not impact performance of the

PowerVault DL2000 – Powered by Symantec media server. Software

compression can however increase end-to-end effective backup

throughput by reducing the size of the backup data before it is

transferred from the backup clients, as demonstrated in the results

above.

Note however that software compression and software encryption

consume processor resources on the backup clients and can

negatively impact end-to-end backup throughput if the clients do not

have sufcient processing power.

Conclusion

The Dell PowerVault ™ DL2000 – Powered by Symantec Backup Exec

provides a high performance, integrated backup-to-disk solution that

ts into any data infrastructure environment.

Further Reading

Additional information about the Dell PowerVault ™ DL2000 –

Powered by Symantec Backup Exec can be found at

http://support.dell.com

www.dell.com/prosupport/mobility

www.dell.com/getprotectedandconnected