Veeam introduced the ability to use Storage Snapshots in v6.5. The first storage product to be supported was the HP StoreVirtual which was closely followed by HP 3PAR StoreServ.
So the question is, it it worthwhile spending the extra 60% per socket on Veeam Backup & Replication Enterprise Plus licensing to be able to use the Backup from Storage Snapshots feature?
Setup
During this test we need to make sure that everything is equal. The vmfocus.com lab has the following VM’s in place for the test:
1 x Domain Controller
1 x SQL Server
1 x vCenter
3 x Blank Windows 2008 R2 servers
All of these servers reside on a single HP StoreVirtual VSA volume provided by two 7.2K 2.5″ HDD in RAID 1.
No workload will be placed on any VM’s between backups.
The target for the backups will be a 5,400 RPM consumer grade 2.5″ SATA HDD in RAID 0 (this is on a different physical host).
The Veeam Server is a Windows 2012 VM which has connectivity into the LAN and iSCSI network
The Veeam Server connects directly to vCenter and is configured to run 4 x jobs in parallel. From a resource perspective it has 4vCPU’s and 8GB RAM.
Veeam Hot Add
OK let’s crack on. For this test I have disabled the iSCSI NIC on the Veeam Server to ensure that the ‘Automatic Proxy Setting’ is left at default for Hot Add. I then created a single backup job called ‘Veeam Hot Add’ which houses 6 x VM’s totaling 138.2 GB in size.
The backup proxy is set to automatic and I have deselected Storage Integration from the Advanced Settings tab.
Both ‘enable application-aware image processing’ and ‘enable guest file indexing’ are NOT selected.
All other settings are left at default.
Veeam Direct SAN Access
Before the test, I rebooted the Veeam Server to make sure all things are equal.
For this test, I created a new backup job called ‘Veeam Direct SAN Access’ and enabled the iSCSI NIC so that Veeam would use Direct SAN Access mode.
Storage Integration from the Advanced Settings tab continues to be deselected.
Veeam Storage Snapshots
Again, before the test, I rebooted the Veeam Server to make sure all things are equal.
For this test I have created a new backup job called ‘Veeam Storage Snapshot’. I have enabled the iSCSI NIC on the Veeam Server and also selected Storage Integration from the Advanced Settings tab.
Every other setting is exactly the same and no data has changed between backups to ensure a true comparison.
Initial Results (Stagnant Data)
I think the results show that all three different methodologies produce fairly similar results in terms of backup performance. With HotAdd being followed by Direct SAN and then Storage Snapshot. However this is with stagnant data e.g. no changes to deltas during backups which doesn’t really happen in most production environments.
| Item |
Hot Add |
Direct SAN Access |
Storage Snapshot |
| VMF-DC01 |
0:26:58 |
0:28:25 |
0:27:39 |
| VMF-FILE01 |
0:24:11 |
0:14:02 |
0:23:37 |
| VMF-SQL01 |
0:29:27 |
0:26:36 |
0:23:01 |
| VMF-VC01 |
0:38:24 |
0:35:46 |
0:36:25 |
| VMF-VIEWC01 |
0:17:36 |
0:16:05 |
0:17:08 |
| VMF-VIEWS01 |
0:15:33 |
0:14:09 |
0:16:23 |
| Overall Processing Rate |
52 MB/s |
51 MB/s |
47 MB/s |
| Overall Duration |
0:40:26 |
0:40:19 |
0:46:29 |
| Bottleneck |
Source |
Source |
Source |
Initial Results (Changing Data)
Like me you are probably thinking, the above ‘Initial Test (Stagnant Data)’ wasn’t really a test as data changes. So to try and simulate a production workload we are going to introduce IOMeter on VMF-FILE01 with the following criteria:
- 8K Block Size
- 100 Random
- 65% Read
- 35% Write
Veeam backup starts and then 1 minute later IOMeter starts for a duration of 5 minutes.
Veeam Hot Add
Veeam Direct SAN
Veeam Storage Snapshots
Initial Results (Changing Data)
| Item |
Hot Add |
Direct SAN Access |
Storage Snapshot |
| Processing Rate |
33 MB/s |
29 MB/s |
27 MB/s |
| Duration |
0:18:36 |
0:18:29 |
0:10:54 |
| Bottleneck |
Source |
Source |
Source |
My Thoughts
In a production environment you are likely to have a mixture of virtual machines, some which are hardly touched and others which are accessed all the time during day and night.
Storage Snapshots are the least intrusive to your production environment and perform better with changing data reducing your backup window by up to 64%. Whereas Hot Add performs best when data is stagnant.
For me, it makes sense to recommend using Veeam Backup & Replication Enterprise Plus for customers who have either a HP StoreVirtual or HP 3PAR StoreServ as they have the ability to choose how they want their backups to work.
Disclaimer
Even though these tests are produced in the vmfocus.com lab, I wouldn’t use the results for a business environment for production workloads.
I would recommend testing Veeam Backup & Replication using the 30 day free trial in your own environment and testing each backup mechanism to understand which one works best for your environment.
VMFocus 