Dell PowerVault DL2200 CommVault - Manual (Page 3)

DELL POWER SOLUTIONS | September 2009 48
Feature Section:
Storage eFFiciency
Reprinted from Dell Power Solutions, September 2009. Copyright © 2009 Dell Inc. All rights reserved.
every backup is almost certain to exceed
operating windows. In these environments,
multi-streaming is virtually a necessity for
concurrent backups.
Uncontrolled concurrency, however,
presents a different set of challenges.
Simultaneous backup activity kicks off a
period of intense processing, coordina-
tion, and movement within a confined
operating window. In a virtualized envi-
ronment, the shared operations and
resources yield a large number of vari-
ables and conditions that can contribute
to failure, rendering the diagnostic pro-
cess long and tedious and making concur-
rency a critical unknown factor.
Figure 1 illustrates the results of run-
ning a set of backup jobs over 10 indi-
vidual VMs using a single backup policy,
highlighting the operational challenges of
consolidated backup policies. The profile
on the left, a traditional file system backup
policy extended to support full VCB back-
ups, performs sequential processing of
each VM and therefore exhibits a much
longer overall backup window than the
profile on the right. Although the method
on the left can help reduce the amount of
concentrated load on the system, even
underutilizing available resources on the
physical server and backup targets, it also
risks missing backup windows and expos-
ing the environment to data loss.
The profile on the right uses multi-
streaming to support the same set of full
VCB backups in significantly less time.
Because ungoverned parallel jobs can
overload the infrastructure and cause
system errors, the UVSA offers two con-
trols to mitigate this problem and help
ensure optimal performance. First, admin-
istrators can set the number of VMs that
can be processed concurrently within a
given policy. Second, the UVSA includes
a built-in time delay between successive
VCB snapshot calls, which helps limit the
impact on the system. As each VM backup
completes, resources are reapplied to the
next VM in the policy. This approach
imposes an operational governor across
the overall process, helping minimize the
impact of unknown conditions through
the standard Simpana job management
resiliency and job restart features.
optimizing Data tranSFer
anD retention
Environments using VCB typically utilize the
image-level mode with granular recovery
options. As more servers are converted to
VMs, therefore, the total volume of nightly
data transfer grows in full VM size incre-
ments, which can have a major impact on
scalability. For example, under a typical
schedule of weekly full backups and daily
incremental backups, 50 traditional servers
averaging 20 GB of data each with a 10 per-
cent daily incremental change rate would
require 1,600 GB of data transfer each week.
If these servers were converted to VMs in
an environment that supports only full VM
image backups, the need to back up all 50
VM images on a daily basis would result in
7,000 GB of data transfer each week—an
increase of 430 percent. In many environ-
ments, this massive inflation can quickly
break the networks and storage budgets.
The BLI capabilities provided by the
CommVault Simpana UVSA allow admin-
istrators to back up only the changed sec-
tions of a VM, enabling them to continue
following weekly full backup cycles as
physical servers are converted to VMs
and helping to limit nightly data volumes
to comparable pre-virtualization levels.
Supporting continued use of existing infra-
structure helps limit the need to invest in
additional tier 2 disks or tape drives as part
of the virtualization project.
Figure 2 illustrates the advantages of
BLI backups and the trade-offs imposed
when such features are lacking. As with the
previous example, this environment has
50 VMs averaging 20 GB each, for a total
of 1 TB of data. A typical B2D profile would
include two to three times this production
disk space for backups and a 30-day disk-
based recovery point objective (RPO), or
30-day data retention; reflecting this typi-
cal profile, administrators have provisioned
2–3 TB of B2D disk space.
The workflow on the left reflects legacy
backup software that does not support
incremental VM backups, requiring daily
full backups of each VM. In addition to
stretching the limits of data transfer capac-
ity each day, this example would fit only
two days’ worth of backups on the B2D
disks—2 TB for two days of full backups,
and 0.5 TB of room to recover granular
file-level data from previous backups (a
typical amount depending on VM size).
Meeting the 30-day RPO would typically
force administrators to store data older
than two days on tape, add disk capacity,
or invest in a deduplication appliance to
help reduce storage requirements.
CommVault Simpana UVSA
• Built-in multiple readers, throttling controls,
and parallel multi-streaming help maximize
throughput while avoiding overload
Legacy backup software
• Lack of multi-streaming capability requires
each VM backup to run sequentially
• Underutilization of VCB proxy server
I/O capacity slows backups, exceeding
the backup window
• Lack of tuning controls can lead
to suboptimal performance and
infrastructure overload
VM 1
VM 2
VM 3
VM 4
VM 5
VM 6
VM 7
VM 8
VM 9
VM 10
VM 1
VM 2
VM 3
VM 4
VM 5
VM 6
VM 7
VM 8
VM 9
VM 10
Maximum backup window Exceeds backup
window
• Ability to add new VMs
automatically without disrupting
backup schedules helps simplify
backup processes
• Ability to restart helps ensure that
interrupted backups do not threaten
data protection
Figure 1. Multi-streaming helps accelerate backups and avoid infrastructure overload