HPC and Public Cloud

Advantages of Deploying HPC in the Public Cloud

Traditional HPC encounters the following issues:

Investment cost is high, expansion is complex, and reusing existing investment is difficult.
Applications are complicated, resource prediction is difficult, flexibility is poor, and efficiency needs to be improved.
Decision making is slow, causing enterprises to lose the market share and opportunities to develop research results.
The application calculation amount is increasing rapidly, posing a higher requirement for performance.

Using HPC in public cloud helps make full use of the following advantages of cloud services:

Reduced TCO
Allows customers to pay as they use, reducing costs and lowering the threshold for small and medium-sized enterprises to use HPC.
Improved efficiency
Supports on-demand provisioning as well as quick deployment and capacity expansion, shortening the time to market (TTM) and scientific research period for products.
Flexible use
- Presets the MPI library, compilation library, and optimized configuration in the image template to accelerate environment deployment.
- Allows enterprise branches and R&D organizations in different geographical locations to coordinate with each other easily, thereby improving efficiency.
- Allows customers to use the cross-region capabilities of the public cloud to share computing resources and massive data, and analyze big data in the cloud.
Optimized performance
- Performance is 30% higher than common ECSs.
- According to test reports, with virtualization optimization measures such as SR-IOV and PCI passthrough adopted in large-scale cloudification, HPC performance is not deteriorated significantly.

Relationship Between HPC and Cloud Services

**Table 1** Cloud services
Cloud Service	Function
ECS	Used for creating high-performance ECSs on the public cloud platform.
VPC	All ECSs in HPC scenarios are located in the same VPC. They are isolated for network security using VPC subnets and network groups.
Image Management Service (IMS)	Image files are required for creating high-performance ECSs. High-performance ECSs are used to create private images.
Elastic Volume Service (EVS)	All ECSs used in HPC scenarios have EVS disks attached.
Bare Metal Server (BMS)	A BMS is a physical server dedicated for individual tenants. It provides remarkable computing performance and stability for running key applications.
Object Storage Service (OBS)	The Object Storage Service (OBS) is an object-based mass cloud storage service. It provides massive, low-cost, highly reliable, and secure data storage capabilities.
Scalable File Service (SFS)	SFS provides ECSs with hosted shared file storage that complies with the standard file protocol (NFS) and that can scale up to the PB level to support massive data and high-bandwidth applications.
Data Express Service (DES)	DES is a massive data transmission service that uses the physical storage medium (USB or eSATA interface) to transmit a large amount of data to Huawei public cloud. It resolves the high cost and long time consumed by massive data transmission over the Internet.
Direct Connect	Direct Connect is a service that allows you to establish a dedicated network connection from your data center to the Huawei public cloud. It enables you to take full advantage of the strengths of public cloud services while continue to use existing IT facilities, and establish a scalable hybrid cloud computing environment.
Cloud Eye (CES)	CES provides a comprehensive monitoring platform for resources such as the ECS and bandwidth. It enables you to monitor alarms, receive notifications, and view reports and visuals in real time to obtain the status of service resources.
Workspace	Workspace provides virtual desktops and applications and allows you to use cloud desktops anywhere and anytime. It provides professional office applications and a more simplified and secure IT office system with higher service efficiency and lower maintenance cost.