Scalability and Infrastructure Sizing
- You can use NAS, SAN, or Directly Attached Storage device to store your backup data. There are no limitations with any hardware vendor.
We do not recommend any specific hardware vendor.
- The performance of the backup job depends on the IOPS. So, you will get better backup performance, if you have higher IOPS.
Note: To calculate the required storage space for your environment, use the Vembu Storage Calculator.
- The estimates provided below are not ‘stretch estimates’ they are safe estimates and are more pessimistic than optimistic. Typical bottlenecks you should look out for are:
- Bandwidth bottlenecks
- Slowing down of data transfer due to low-performance switches/routers
- The hard disk write speeds at the storage targets
- CPU utilization by other non-related processes running on the Vembu BDR backup server
Backup Environment
Vembu BDR Backup Server Configuration
OS: Windows 2012 R2 DC (Physical Machine)
RAM: 16 GB
CPU: Intel XEON CPU 2.10 GHz (4 cores)
Network: 1 Gbps
Vembu BDR Version: v4.0.0
Hypervisor: VMware vSphere
No. of VMs: 1
VM data: 2 TB
Backup Type
|
VM Data Size
|
CPU Utilization
|
Memory Consumed
|
Total time taken to complete the backup
|
Transfer rate
|
Full Backup
|
2 TB
|
22 %
|
3145 MB
|
383 Mins
|
803 Mbps
|
Incremental
|
55 GB
|
16.5 %
|
401 MB
|
11 Mins
|
751 Mbps
|
Note: The above measurements are taken for one backup job with one VM. If multiple backup jobs are active then it will share the memory and CPU. As stated earlier, each backup job will approximately require 500 MB RAM.
- It is important to note that there are no inherent scalability restrictions for Vembu BDR backup server when it comes to the maximum number of simultaneous backups possible.
- If a larger number of backup jobs are simultaneously configured to the Vembu BDR backup server, then the only effect will be a corresponding degradation of performance in terms of time taken to complete the backup for all the backup jobs.
- The maximum number of backup jobs that can be supported by a single Vembu BDR backup server depends on -
- The bandwidth of the network used
- Time taken for each backup job
- For example, the above illustration assumes that backups are typically scheduled during ‘non-office hours’ (hence the 10-hour window); this is not always the case. Especially, if the backups can happen non-intrusively in the background while you do your regular work.
Storage Device Throughput And Concurrent Writes:
- The throughput of storage devices (IDE/SCSI/SAS/NAS/SAN) is normally benchmarked based on maximum throughput achieved while performing sequential writes into the device.
- When multiple VMs are backed up simultaneously to the Vembu BDR backup server, then the server will concurrently write different files for different VMs onto the storage device.
Note: When concurrent writes from multiple threads are being done, the throughput of the storage device determines the performance of Vembu BDR backup server
- The reason storage devices do not perform well when multiple threads are writing to it concurrently is that the I/O seek that has to be done between writes from different threads can slow down the performance significantly.
- Hence, it is imperative that the storage device used to backup the data is of the highest quality and has the ability to scale and perform well when 100s of different threads write to the storage device concurrently.