.){kind=link}
Key Takeaways:
– Microsoft introduces Azure confidential virtual machines (VMs—NCC H100 v5 SKU) with enhanced security and performance attributes.
– These VMs combine data protection from AMD EPYC processors with high performance from NVIDIA Tensor Core GPUs.
– Azure’s confidential computing offers customers increased flexibility and security for various workloads.
– The performance of these VMs is ideally suited for tasks such as inferencing, fine-tuning, and training models.
– NCC H100 v5 VM SKUs offer a hardware-based Trusted Execution Environment (TEE) that improves the security of VMs.
– Azure NCC H100 v5 GPU VMs are currently available only in East US2 and West Europe locations.
Security and Performance Expectations
Microsoft has stepped up its game with the announcement of the general availability of Azure confidential virtual machines (VMs—NCC H100 v5 SKU). These VMs provide an extra layer of protection, combining the sophisticated data protection from 4th-generation AMD EPYC processors with high-performance NVIDIA Tensor Core GPUs.
Following a preview of these VMs last year, confidential computing on GPUs in Azure offers customers more options and flexibility for managing their cloud-based workloads securely. This upgrade is not just about enhanced security, the performance of these VMs also makes them perfect for tasks such as inferencing, fine-tuning, and training small to medium-sized models.
Broad Range of Models
The Azure VMs are compatible with various models including Whisper, Stable Diffusion, SDXL, SSD, Zephyr, Falcon, GPT-2, MPT, Llama2, Wizard, and Xwin. Choosing these VMs allows customers to handle different projects with diverse model requirements without worrying about compatibility issues.
Enhanced Security with TEE
The NCC H100 v5 VM SKUs feature a hardware-based Trusted Execution Environment (TEE), an addition aimed at improving the security of guest virtual machines. The TEE is a protection barrier against unauthorized access to VM memory, flags any potential access by hypervisor and host management code, and confirms the non-existence of any illicit operator access.
In addition, customers can initiate attestation requests inside these VMs. This will assure them that their data is secured in a properly configured TEE before releasing keys and starting sensitive applications. Verifying the TEE’s configuration is crucial to avoiding breaches and unauthorized data access.
Real-world Applications of Azure’s New VMs
Several businesses have already leveraged Azure NCC H100 v5 GPU VMs for a variety of purposes. The applications range from confidential audio-to-text inference, data privacy with confidential computing, to video analysis for incident prevention.
The automotive sector stands as a notable example. There, stable diffusion projects involving sensitive design data are managed using these VMs. This demonstrates the flexibility and wide application range of Azure’s new offering.
Competitive Landscape
Microsoft is not the only cloud-service provider making use of NVIDIA’s powerful H100 Tensor Core GPUs. AWS and Google, two other big hyperscalers, also offer the technology. For example, AWS provides H100 GPUs with EC2 P5 instances, optimized for high-performance computing and AI applications.
Performance Boost with NVIDIA’s H100
NVIDIA’s H100 Tensor Core GPU, built on the Hopper architecture, promises to significantly boost performance in large-scale AI and HPC workloads. Azure’s NCC H100 v5 VMs leverage this tech to deliver superior performance all while ensuring strong scaling for AI and HPC workloads.
NCC H100 v5 Availability
Currently, Azure’s NCC H100 v5 virtual machines are only available in two regions: East US2 and West Europe. Nonetheless, with the mounting demand for secure, high-performance cloud solutions, Microsoft might consider extending the availability to other regions.
In conclusion, Microsoft’s announcement of the GA release of Azure NCC H100 v5 SKUs is a significant step towards ensuring security and performance in cloud-based solutions. The inclusion of NVIDIA Tensor Core GPUs coupled with AMD EPYC processors directly addresses customer concerns around data protection and performance in the cloud. The wide range of applications also highlights the versatility of these VMs in managing diverse workloads.