First, let’s review the most common use cases of backup based on this solution.
The base use case is identical to the one with Acronis solution:
- Backup from the client’s stage to #CloudMTS cloud storage;
- Free data migration to #CloudMTS cloud.
There are unique scenarios for backup services based on Veeam solution.
Virtual machines backup inside #CloudMTS cloud
In this case, Veeam Backup and Replication integrate with VMware vCloud Director IaaS service from #CloudMTS. This allows Veeam to backup virtual machines hosted on Elastic Cloud.
As part of this use case, the client is provided with access to a self-service web portal that is integrated with vCloud Director. This way the user will have a single sign-on. User himself can create and run tasks for VM backup (including those transactionally consistent for applications that support VSS technology), as well as restore virtual machines (or individual files) located inside #CloudMTS cloud.
Managing backup copies in the protected segment FL#152RU
In this use case, the user can store backup copies of data that contain personal information in the protected segment that runs in compliance with FL#152RU.
This is achieved through the joint use of Veeam agents installed on the backup object, in combination with S-Terra crypto gate, which also functions as part of client’s infrastructure and provides secure transfer of backup to the FL#152RU segment.
If client’s infrastructure already has a Veeam Backup & Replication server, then it is possible to replicate virtual machines 2 ways:
- Replication of virtual machines from the client’s stage to Elastic Cloud. Many confuse this scenario with DRaaS service from #CloudMTS. In cases when Veeam solution is used, client replicates virtual machines to pre-allocated Elastic Cloud resources with the help of Veeam Cloud Connect Replication. The replica is not active until a contingency occurs on client’s main stage or there is a need to launch it on the backup site. After that, user independently activates the virtual machine through vCloud Director and runs copies of his servers in Elastic Cloud as of their last replication.
- Offsite Backup – a scenario when the created backups are added and stored in #CloudMTS repository.
Veeam Backup & Replication Server Support at Client’s stage
Professional & Managed services provide an option for clients for deployment, configuration and support of Veeam Backup & Replication server on their own stage performed by #CloudMTS experts.
So, let’s review one of our use cases.
We needed to expand one of the current Veeam repositories to the maximum possible capacity, at minimal cost, with minimal downtime, and in compliance with corporate standards. Currently, repositories are traditionally hosted on Huawei 5288V5 servers configuration 2 IntelXeon Gold 6132, 256Gb RAM with SAS Expander via SAS3508 with 36 disks.
We agreed on a downtime of 15 minutes, mounted into the server an Avago MegaRAIDSAS 9380-8e controller + CacheVault LSICVM02 connected to PCI-Express 3.0 x8, started the server, checked the firmware/drivers and returned to commercial use.
The server operated as usual, at that time, we engaged in the installation and switching of one of the WD/HGST 1ES0352 SE4U60. racks.
The installation was carried out quickly and we moved on to software
MegaRAID Storage manager is launched and we begin to configure RAID60. This configuration has its own pros and cons. We need protection for the time of rebuild (reconfiguration) from the fan failure of capacious disks, which, as a rule, come from one batch, and the maximum possible performance of read and write sequential operations.
As a result, we have RAID60 4 SPAN with 14 disks + 4 disks in spare per total capacity, without the need to splice disks at the OS level, and that provide protection from failure of 8 disks, 2 in each SPAN. The rest is up to statistics and luck.
We turn the derived virtual disk into a logical drive and format it in ReFS with a 64KB block.
DISKPART> list volume
Volume ### Ltr Label Fs Type Size Status Info
---------- --- ----------- ----- ---------- ------- --------- --------
Volume 5 F VD_0 ReFS Partition 436 TB Healthy
>fsutil fsinfo refsinfo F:\
REFS Volume Serial Number : 0x583498f53498d6fa
REFS Version : 3.1
Number Sectors : 0x000000da48fa0000
Total Clusters : 0x00000001b491f400
Free Clusters : 0x00000001b383cff8
Total Reserved : 0x0000000000800852
Bytes Per Sector : 512
Bytes Per Physical Sector : 4096
Bytes Per Cluster : 65536
Checksum Type: CHECKSUM_TYPE_NONE
What about the performance?
The block that is responsible in Veeam for read and write operations from storage equals to 256KB or 512KB, depending on the type of operation.
We are interested in consecutive read and write operations.
When deduplication is used (i.e. removal of duplicate parts, dubbed versions), the number of consecutive operations will undoubtedly be different. However, we are interested in the maximum performance possible without deduplication, since not all of our repositories are subject to deduplication.
Load will be created with IOMeter in 16 workers with 36 outstanding I/Os each. We start the process of filling up the storage and forget about it for a couple of weeks. Filling is carried out via 64KB block.
The performance output by disk can be monitored in the task manager. We can now review the load in the task manager, not in detail, but enough to understand whether the process is “live” and whether the disk is full.
CPU load is about 10%. 9% from which are consumed by Dinamo IOMeter
The values obtained are as much as plus or minus 5.8-6.3 GB/s, which is close to the standard bus value in PCI-Express 3.0 x8 of 7.88 GB/s. It’s a pity that the current configuration of the server and its almost complete utilization on PCI-Express slots does not allow to use at least two SAS controllers.
Further, we perform consecutive read and write operations with a 256KB and 512KB blocks in a proportion of 25% for each type of operation with full capacity of the derived local disk. Therefore we have the following output:
As a result, we have aggregate throughput of 6.2 GB/s with excellent response of 21ms at the local disk level.
Opportunities for further drive expansion
As we noted above, our server is already fully utilized what comes to PCI-Express ports and it is not possible to install additional SAS controllers on it. One question immediately comes to mind: what can be done in this case?
It turns out it's easier than we thought. The WD/HGST 1ES0352 SE4U60 product line allows to cascade up to 4 racks, i.e. reach 2.16Pb RAW capacity, while maintaining the received bandwidth.
As a result, we have a nice configuration, with the ability to quickly and painlessly expand one of Veeam repositories. Together with total RAW rack capacity of 540Tb and with 14 disks and 4 disks in reserve after the formation of RAID60 4SPAN, we received a loss in capacity of about 100Tb and excellent performance.
Text prepared by: Dmitry Sudarikov, lead architect, Sergey Sklabovsky, product manager, Anton Kalinkin, head of Professional and Managed Services, and Sergey Shashin, product manager.