Oracle Japan recently held a briefing session on “Oracle Cloud Infrastructure HPC” and “Oracle Observability and Management Platform”.
Recently, there has been an increasing tendency to use high performance computing (HPC) and GPU as cloud services, and major cloud providers are providing related services. Under such circumstances, what are the strengths of Oracle’s HPC service?
Kota Kondo, Senior Manager, Business Promotion Division, Technology Business Strategy, explained that HPC and GPU cloud utilization is progressing in various industries such as manufacturing, finance, life sciences, oil & gas, weather, media, and games. did.
Because HPC is suitable for the cloud, he said. Regarding the reason, Mr. Kondo said, “You can use cloud services as much as you want. Therefore, you can use a large amount of resources before the delivery date and stop charging when you finish using it. In addition, you can use the latest technology.” Said.
In other words, by using HPC in the cloud, it is possible to shorten product development time and cost, and at the same time, it is possible to use the optimum technology according to the application when needed. It’s not so easy to have the latest HPC-related technologies in-house.
According to the history of Oracle’s HPC, the personnel of Sun Microsystems, which was acquired in 2010, are involved in the development. We introduced a bare metal server in 2016, and “this has a positive impact on the performance of our HPC,” Kondo said.
Oracle Cloud Infrastructure for HPC is said to deliver performance equal to or better than on-premises InfiniBand HPC clusters than low-latency networks and bare metal servers for HPC. Mr. Kondo showed the results of a benchmark comparing the delay of the company’s network with Amazon Web Services and Google Cloud, and emphasized the company’s superiority.
The “Oracle Cloud Infrastructure HPC” lineup consists of three services. The CPU (AMD, Intel) and GPU (NVDIA) provided by each are different. On September 30th, instances of the NVIDIA A100 Tensor core GPU were available. An instance of Intel Ice Lake and an instance of AMD Milan will be available in the first half of 2021. In addition, a new instance, the Ampere Altra Processor, will be available in the first half of 2021.
Next, Ms. Mikako Shimizu, Manager, Business Promotion Division, Technology Business Strategy, explained about “Oracle Observability and Management Platform”. The product is a group of operation management services announced in October this year. It provides visibility and control of software and the entire software stack on-premises in a multi-cloud environment, automatically detects anomalies with machine learning and makes corrections in near real time.
In IT operation management, Mr. Shimizu said, “While multi-cloud and hybrid environments are being introduced in earnest, system operations are becoming more complicated.” “Efficiency is reduced by using multiple individually optimized tools.” “Inhibition of DevOps system development due to lack of cooperation between developers and administrators” was mentioned, and the product will solve these problems.
The service includes existing services such as Monitoring, Notifications, Events, Functions, Streaming and OS Management, as well as the newly announced Logging, Logging Analytics, Database Management, Application Performance Monitoring, Operations Insights and Service Connector Hub. ..
“Logging” is a platform that facilitates log capture / management / analysis. Enable full-asset logging with a single click to securely manage all log types for audits, infrastructure, databases, and applications. Built with open standard technology, it leverages fluentd for log collection and is CNCF cloudevents 1.0 compliant.
“Logging Analytics” applies machine learning to discover patterns and outliers in logs to identify system anomalies. Logs can be archived and recalled based on user-defined policies.
Operations Insights provides insights into future resource demand and application performance. Specifically, it analyzes and predicts the usage status of database resources based on long-term historical data, identifies servers with high usage rates and servers with low usage rates, and optimizes operating costs.
“Service Connector Hub” facilitates data linkage between OCI and third-party services. Specifically, “Easy to move data to a third-party observable tool using Kafka-compatible OCI Streaming service” “Easy sending log metrics to OCI Monitoring” “One to object storage” “Click to archive logs” “Automatic repair and alert notification of log events using OCI Functions and Notifications” is possible.
The above four services are currently offered in the Tokyo region. The following two services will be released in the future.
“Application Performance Monitoring” visualizes application performance and user experience and derives satisfaction. Measure end-user performance from browser to application and analyze performance, availability, and load using server browser transaction metrics.
Database Management is an Oracle Database on-premises and across the cloud
Resist an integrated view for monitoring and managing, and centrally manage the usage of multiple databases in a hybrid environment.
Users of “Oracle Cloud Infrastructure” and “Oracle Dedicated Region Cloud @ Customer” can use the service immediately.