Harnessing HPE Alletra Storage Server 4000 and Veeam for unparalleled data protection

[StorageExperts]

The joint solution engineered by HPE and Veeam is a secure, scalable, and cost-effective all-HDD data protection solution. It not only meets but exceeds expectations in backup and restore performance. Learn how.  – By Grishma H N, Solutions Engineer, HPE

Setting new benchmarks

In the high-stakes world of Formula 1, every millisecond counts. A well-tuned, powerful car combined with a skilled driver can make the difference between standing on the podium and being just another contender. Similarly, the performance and reliability of your storage infrastructure are crucial for protecting your data. As data volumes explode, the need for a fast, manageable backup solution becomes more critical.

Recall how we at HPE, in partnership with Veeam, set high backup/restore benchmarks by integrating HPE Apollo 4510 server with Veeam Availability Suite. With the new generation of HPE Alletra 4000 servers and HPE ProLiant DL380 Gen11 servers, we’ve pushed those limits even higher. Our latest tests with Veeam Data Platform 12 not only exceed the previous standards but also provide a jaw-dropping backup performance of 58 TiB/hr with up to 1.6 PB of native capacity in a single server with only SAS HDDs. This solution isn't just about high performance. It offers a secure, scalable, and affordable data protection infrastructure for any workload.

Let me take you through a short exploration of the solution design and its performance to uncover its effectiveness without any superficial claims.

Solution overview

The data protection solution featuring Veeam Backup & Replication 12.1 leverages the high throughput and high capacity of HPE Alletra Storage Server 4140, HPE Alletra Storage Server 4120, and HPE  ProLiant DL380 Gen11 Server as backup targets. Each server is tuned and validated for different performance and capacity needs. They can be utilized either as a Hardened Linux repository server for heightened security against ransomware attacks or as an all-in-one server for simplified management. Equipped with high-density compute and storage within the same chassis, these servers adeptly support all components of the Veeam backup infrastructure. You can start with a single node and scale- up or scale-out by adding more nodes for a larger solution.

Understanding the benchmark test environment

The HPE Alletra Storage Server 4140, HPE Alletra Storage Server 4120, and  HPE ProLiant DL380  Gen11 servers were rigorously tested and optimized for performance. Here are the details of the test environment used for benchmarking the large model, HPE Alletra Storage Server 4140.

Protected workload type

The test data source consisted of 60 VMs spread across 8 ESXi servers. Each VM had a dataset between 200 GB and 300 GB. To reflect a real production environment, the VMDK of the VMs were loaded with randomly generated binary data with an exact compression ratio of 2:1 so that after the backup job completes, backup file will be half the size of the data loaded into the VMDKs.

About 80% of the VMs were on volumes across the three HPE primary storage arrays: HPE GreenLake for Block Storage built on HPE Alletra Storage MP, HPE Alletra 9000, and HPE Alletra 6000. These primary storage arrays are fully integrated with Veeam Backup & Replication, thus supporting the most advanced Veeam Snapshot Orchestration features. The rest of the VMs were on the internal NVMe disks of the ESXi servers. Each ESXi server also hosted 2 Windows Veeam Proxy VMs for backing up the source VMs via Hot-Add backup mode.

Backup target

The HPE Alletra Storage Server 4140 is a 4U server available in two configurations: 68 disks and 92 disks. The server used for benchmarking was configured with these components.

• CPU: 2* Intel Xeon-Gold 2.0 GHz 24 core processors
• Memory: 16 * 16 GB DIMMs
• HDD for Backup Storage: 68 * 20TB disks
• SSD for OS: 2 * 480 M.2 NVMe drives
• Hardware RAID controllers: 2 * HPE SR932i-p storage controllers with 32 SAS lanes and 8GB wide persistent cache
• Fibre Channel (FC): 2 * 32 Gb HBA card for FC connectivity with HPE Primary Storage arrays. This is required for all-in-one configurations with hardware snapshot integrated backups.
• LAN: 2 * 100 GbE ports that were teamed in active/active load balancing mode for high availability.

As the server has many disks, using two storage controllers was found to be necessary. The 68 disks were grouped in 2 RAID 60 volumes, each volume formatted with XFS File System (with block cloning). Later, the 2 File Systems were used as performance extents of an immutable Veeam Scale-Out Backup Repository with a total usable capacity of 1.08 PiB!

A backup job was created using this Hardened Scale-Out Backup Repository with Storage settings as shown in the below screenshot.

 Performance highlights

The screenshot below shows the maximum multi-stream backup throughput when the backup job was in progress.

The solution with HPE Alletra Storage Server 4140 produces a remarkably steady backup throughput of 16 GiB/s average and a peak throughput of 19 GiB/s. Veeam Data Platform is NUMA aware which maximizes the system’s parallel processing capabilities with low latency. Additionally, Veeam can directly write the backup data to the disk without utilizing the OS buffer cache and hence the throughput that you see is not influenced by the RAM buffer.

In the above screenshot, the job is 68% complete, and 26 out of 56 VMs with approximately 13 TB of data were backed up in a span of just 15 minutes. The average throughput dropped as the job progressed and few VMs were left to be backed up. In a real production environment where there are hundreds of VMs to be backed up in batches, the backup processes for each kick in one after the other, and this high throughput of 16 GiB/s can be sustained till the end of the job.

The system was monitored continuously during the job run to ensure that CPU and memory utilization by the Veeam backup processes was not overloading the system. Thanks to the perfectly chosen components/parameters for tuning the systems and performance optimization measures in the Veeam Data Platform, the CPU utilization remained well below 20%, and the system had sufficient free memory while maintaining a tremendous write throughput.

The server is significantly fast on restores, too. We found that the primary obstacle to performance was never the HPE Alletra Storage Server 4140.  The bottleneck was in fact the mock-production environment or the Veeam proxy VMs/servers used for testing.  So, we had to bloat the source environment and restore 60 VMs to 8 ESXi servers concurrently using 16 hot add proxy VMs spread across the ESXi servers. The backups of 60 VMs were evenly spread across the two RAID 60 volumes on HPE Alletra Storage Server 4140. Screenshots below show the iostat output captured during the restore process for both the repository volumes that were used as performance extents in the Veeam Hardened Scale-Out Backup repository. 

The system was able to deliver a high read throughput of approximately 3.25 GiB/s on average per RAID 60 repository volume during concurrent restore. The read latency (“r_await”) was less than 100 ms (30 ms in this case). Anything above 100ms was considered to be too much. As the VM backup data is 2:1 compressible, this read throughput translates to a maximum restore throughput of approximately, 6.5 GiB/s per XFS file system of the Scale-Out Backup Repository. So, the two XFS file systems consuming the entire capacity of the HPE Alletra Storage Server 4140 deliver a restore throughput of 13 GiB/s.

A similar testing methodology was used for benchmarking the HPE Alletra Storage Server 4120 and HPE ProLiant DL380 Gen11 Server. Each server delivered impressive backup performance with minimal tuning.

 Performance that can be attained varies by environment. However, the benchmark tests demonstrate that it is very unlikely the HPE Alletra Storage Server 4140/4120 or HPE ProLiant DL380 Gen11 Server will be the bottleneck.

Take your data protection to the next level of ultimate performance

The lab tests clearly demonstrate that the data protection solution with Veeam Data Platform and HPE Alletra Storage Servers and HPE ProLiant DL380 Gen11 Servers is ultra-fast. With other solutions, there’s a trade-off when it comes to security, speed, and cost. With this joint solution by HPE and Veeam, the demands of today's data-driven environments are met. The solution provides essential security, easy manageability, and quick recoverability, particularly crucial in the face of increasing ransomware threats.

The solution offers remarkable flexibility and scalability, efficiently supporting both small businesses and large enterprises without compromising performance. For those interested in the technical details and our comprehensive testing methods, we encourage you to consult our detailed technical paper.

To learn more

Watch the video on protecting your data with HPE Alletra Storage Server 4140 and Veeam Data Platform.

And check out:

• Solution brief: HPE all-in-one data protection storage solutions with Veeam
• Technical paper: Veeam Backup & Replication on  HPE Alletra Storage Server 4120/4140 and HPE ProLiant DL380 Gen11 Server

---

Meet Storage Experts blogger Grishma H N, Solutions Engineer, HPE

Grishma is a Worldwide Hybrid Cloud Solutions Engineer at HPE. She began her IT career at HPE R&D as a systems software engineer and currently focuses on designing and championing data protection solutions for HPE. Connect with Grishma on LinkedIn.

Storage Experts
Hewlett Packard Enterprise

twitter.com/HPE_Storage
linkedin.com/showcase/hpestorage/
hpe.com/storage