AMD Benefits for Dell vSAN Ready Nodes
Tue, 15 Feb 2022 16:12:31 -0000
|Read Time: 0 minutes
In a previous blog I explained how the availability of a rich offering of platforms to support vSAN Ready Nodes provides additional value to customers.
Apart from the various form factors including tower, rack, modular, and ruggedized models, or the existence of 1 and 2U offers and compute options ranging from one to four sockets, Dell vSAN Ready Nodes also provides a choice of processor: Intel and AMD.
In this blog, I explore some of the key advantages of the AMD based processor, focusing on:
- Technical specifications
- Performance
- Security
Technical specifications
Dell has four AMD based servers certified as vSAN Ready Nodes: the Dell EMC PowerEdge R6515, R6525, R7515, and R7525.
These four models incorporate the latest generations of AMD EPYC Rome (Series 7002, AMD EPYC 2nd Gen) and Milan (Series 7003, AMD EPYC 3rd Gen) processors in VMware vSAN Ready Nodes All-Flash and Hybrid configurations, as described in the following figure:
The AMD portfolio consists of two one-socket servers (Dell PowerEdge R6515 and R7515) and two two-socket servers (Dell PowerEdge R6525 and R7525).
The Dell PowerEdge Rxxx5 servers include the new AMD EPYC 7003 (Milan) Series Processors, that have up to 64 cores per processor, based on “Zen 3” architecture. They also introduce new per-core cache memory levels (32 MB L3 cache) while continuing to offer the class-leading PCIe® 4 connectivity that defined the EPYC 7002 (Rome) Series CPUs memory bandwidth.
These servers present a vast I/O bandwidth profile with 128 PCIe™ 4.0 Lanes in a Single Socket, up to 160 PCIe 4.0 Lanes in a Dual Socket and 64 GB/s bi-dir bandwidth per link, 512 GB/s per socket.
Single socket configurations are beneficial in terms of cost and energy footprint because they can compete in performance with dual socket configurations, with significant cost and power savings (280W per AMD EPYC 7763 vs 350W in 2x Intel Xeon Platinum 9242, and 64 cores in the AMD EPYC 7763 vs 48 cores in the Intel Xeon Platinum 9242).
Both AMD Rome and Milan processors feature configurable Non-Uniform Memory Access (NUMA), settings that can go from one NUMA node per socket (NPS1) to four NUMA nodes per socket (NPS4). NPS1 configurations create a larger memory domain, with potentially higher latency and throughput, while NPS4 configurations produce smaller memory domains that help to reduce the ESXi scheduler memory latency.
Optimal NPS configurations depend on the user’s vSphere workload. In general, workloads with large block sequential reads benefit from NPS1 configurations, which achieve almost double the throughput of NPS2 and NPS4 at half the latency. Small block random read type workloads show exceptional performance from NPS2 and NPS4 configurations.
NPS settings have much less impact in heavy write workloads. This probably occurs because all writes are consumed by the cache drive before destaging to the capacity drive.
Performance
AMD EPYC processors have demonstrated an excellent performance profile over the years. They have yielded impressive results not only in the VMware space, but in almost any workload test.
For example, the AMD site shares some general numbers with VMware, VDI, database, high performance computing, or pure integer calculation workloads. For more details, see AMD EPYC™ Tech Docs and White Papers.
VMark is a leading performance benchmark in the VMware space. Dell vSAN Ready Nodes (particularly those based in AMD) demonstrated impressive results for the most recent VMark 3.x tests.
Dell vSAN Ready Nodes make up three out of four of the best global scores, measured by the number of VMmark tiles that ready nodes can run.
The remaining ready node of these four used three times more hosts and sockets than the Dell AMD based nodes. The price implications are demonstrated in the following figure.
With 64 tiles and a score of 63.01, the Dell R6525 vSAN Ready Node shines as a solid performance leader in terms of VMware workloads.
AMD based ready nodes also lead the charts for single-socket systems, as shown in the following image:
The top scoring systems were the Dell R7515 and Dell C6525 vSAN Ready Nodes, with 16 tiles and scores of 15.18 and 13.74, respectively.
Security
AMD EPYC processors create a new standard for secure memory encryption (SME) by making it possible to encrypt the contents of the main memory just by changing a setting in the system BIOS. In encrypted memory systems, cold-boot attacks have a low chance of divulging memory contents because all the data is encrypted. High performance encryption engines integrated into the memory channels help improve performance
Second generation EPYC processors have even more increased security, with performance-optimized countermeasures against known attacks.
Specific to virtualized environments, the second generation of AMD EPYC processors have introduced Secure Encrypted Virtualization (SEV). This technology encrypts each virtual machine with a unique key that in known only to the processor, with up to 509 contexts. This improves protection and data confidentiality in virtualized environments. This includes protection for instances of a malicious virtual machine that finds a way into your virtual machines’ memory or a compromised hypervisor that looks inside a guest virtual machine.
This means AMD EPYC processors are less vulnerable to attacks such as Meltdown, Spectre v3a, LazyFPU, and MDS.
With Secure Boot, AMD creates a hardware root of trust, enabling only trusted code to be loaded and run through BIOS load, helping prevent the injection of malicious code prior to the loading of the operating system. This feature is managed by a dedicated security processor (AMD Secure Processor) that lives alongside the CPU cores.
Conclusion
AMD based Dell vSAN Ready Nodes provide tangible benefits for vSAN deployments, have significant performance and security advantages, and are cost-friendly.
You can read more about Dell vSAN Ready Nodes at the Dell Technologies Info Hub Solutions for vSAN Ready Nodes.
Author: Inigo Olcoz
Twitter: @VirtualOlcoz
References
Related Blog Posts
Simplifying Security Operations for Dell HCI Platforms with NSX
Thu, 08 Sep 2022 16:58:04 -0000
|Read Time: 0 minutes
Today, most technology companies in the IT space work to offer customers not only the best technology innovations but also those that help simplify their day-to-day lives.
One example of this is the new vCenter plug-in for NSX-T, introduced with vSphere 7.0 Update 3c and NSX-T 3.2. Through this new deployment method for NSX-T, management and operations users can now use NSX-T as a plug-in for vCenter, similar to how earlier versions of NSX were configured. Through wizard-assisted operations, security policies can easily be configured, deployed, and operated within vCenter.
Figure 1. The new vCenter plug-in for NSX-T simplifies security deployment and operations
For Dell HCI platforms such as VxRail, vSAN Ready Nodes, and PowerEdge servers hosting vSAN-based workloads, NSX becomes an optimal network and security engine.
Figure 2. Dell HCI platforms such as VxRail or vSAN Ready Nodes become the perfect targets for the new vCenter plug-in
The whole process is simple. It can be completed by following these steps:
- Install NSX-T Manager and provide a license key.
- Install the new method to configure and operate NSX security, the vCenter plugin for NSX.
- Configure the distributed firewall policies for the HCI cluster:
a. Define infrastructure services as needed (DNS, DHCP, custom…).
b. Create the environment to consume the defined infrastructure services. - Define how the elements in the environment can communicate with each other.
- Define communication strategies for applications in the environment.
- Review and verify the defined security policies before they are published and effective.
Figure 3. Defined NSX security rules can be reviewed before going live
If you want to learn more about how simple security operations can become with the new vCenter plug-in for NSX, take a look at this video.
Author: Inigo Olcoz
Twitter: VirtualOlcoz
References
- VxRail Info Hub
- vSAN Ready Nodes Info Hub
- HCI Security Simplified: Protecting Dell VxRail with VMware NSX Security
- Simplifying Security Deployment and Operations for Dell HCI Platforms
- Video: Simplifying HCI Security with the New vCenter Plug-in for NSX
The Benefits of Composable Infrastructure for VMware Cloud Foundation on vSAN Ready Nodes
Thu, 25 Aug 2022 17:14:52 -0000
|Read Time: 0 minutes
Discussions comparing the public cloud to on-premise architectures have shifted to discussions about the best way to build a hybrid model.
Multi-cloud delivery services alongside on-premise infrastructure seems to provide the versatility that organizations need for business critical missions, however they are often complex and costly. Ending up with multiple administrative teams, toolkits, and processes doesn’t appeal to most organizations as a winning horse.
As usual, simplicity comes to our rescue. An integrated offering that provides customers with the advantages of both the public cloud and an on-premise infrastructure helps organizations with the following:
- Control infrastructure price and performance
- Improve their deployment times and thus time to value
- Minimize cost and complexity, through a single set of skills and tools
Most organizations that are interested in these benefits are not prepared to build a hybrid cloud themselves. Running an on-premise datacenter or building assets in the main public cloud tends to be more familiar for most IT teams, but managing both at the same time may be beyond the capabilities of some existing IT teams.
In these cases, a trusted partner that brings guidance and innovation to follow this path is crucial. Such a partner can provide a set of familiar management tools with all the administrative and analytics capabilities required to monitor business workloads—such as tools that can be extended to multiple public cloud providers, giving organizations the advantages of compelling cost, efficiency, and speed.
Because most organizations today are using virtualized on-premise infrastructure, VMware comes to my mind as a perfect partner for this journey, having a mature technology offering to help companies build a hybrid cloud.
VMware on Dell offers infrastructure to build a foundational hybrid cloud. This architecture makes the on-premise more cloudy and enables workload mobility to place every workload, whether it is in a public cloud or in our on-premise infrastructure.
Dell PowerEdge MX offers a software defined data center that empowers organizations towards a hybrid cloud model. For example, when this infrastructure is coupled with VMware Cloud Foundation (VCF) and VMware on AWS organizations can build a hybrid cloud using popular VMware tools and capabilities.
Figure 1: MX composable infrastructure chassis
MX chassis is built with capabilities that perfectly match vSAN requirements. Each node can hold six drives, which ensures that that two of the drives are caching devices within the vSAN cluster for optimal performance.
A low latency smart fabric is built into the MX design. This minimizes the risk that lack of throughput or excessive latency presents for a hybrid cloud deployment. Expandable and extensible by design, the MX family can deliver cloud value across many technology generations.
Dell MX management is integrated with VMware Cloud Foundation 4.x. The MX hardware has the systems’ management and APIs to hook into the VMware consoles that customers are used to. This way we can deploy VCF into the MX infrastructure though a simplified path that eliminates tedious stepwise processes around setup, monitoring, provisioning, and management.
Dell and VMware have also worked together to improve telemetry and insight, giving operators an improved view of resource utilization for each node and chassis in the MX infrastructure.
In summary, the advantages of running VCF on Dell vSAN Ready Nodes based on PowerEdge MX servers include:
- A scalable network design that provides a low latency, automated and not oversubscribed fabric to support any workload deployment the business may need
- A secured stack all the way from the hardware to the VMware management console
- An integrated management with tool set that helps IT admins provision, monitor, and maintain their MX based vSAN Ready Node farm. One console can manage multiple chassis and even other server and storage types.
Figure 2: MX integrated management simplifies hybrid cloud operations
One way to see the integration of VCF with MX based vSAN Ready Nodes in action is to deploy a new workload cluster of Dell vSAN Ready Nodes MX750c into an existing VCF infrastructure.
Principled Technologies performed this test and showed how simple and straightforward the expansion process is with MX servers. They also showed, by following the same expansion process with two generations of MX servers (MX740c and 750c), that the tools and processes remain the same across server generations.
The engineers completed the expansion process in just two hours and 21 minutes. It took the same time for each of the two server generations.
For more information about this test case, see the Principled Technologies report.
Conclusion
Although building a functioning hybrid cloud to support real business workloads may look complex, it can be simplified if by using the proper technologies and tools. VCF combined with Dell vSAN Ready Nodes MX composable infrastructure offers a perfect duo to reach an organization’s hybrid cloud goals.
You can read more about Dell vSAN Ready Nodes at the Dell Technologies Info Hub.
Author information
Iñigo Olcoz
Twitter: @virtualOlcoz