CloudLine Metrocluster

Main articles:

• What is it IaaS
• Virtualization. Classification and applications

2021

Launch of cloud services based on Rostelecom data center in Novosibirsk

The company DataLine on August 17, 2021 announced the launch of its own cloud services in Siberia. clouds The DataLine availability zone will appear based on "" data center in.Rostelecom-Data Center Novosibirsk

Data Center customers will now Novosibirsk be able to leverage virtual failsafe-based resources clouds to VMware host business-critical applications and. data Thus, the accessibility zone expands the geography of the cloud DataLine up to 5 sites, including NORD OST the To Moscow data center St. Petersburg in and MegaData Center "Udomlya" in. Tver region

In addition to infrastructure as a service (IaaS), a wide range of databases provided according to the DBaaS model and ready-made Kubernetes managed clusters for working with containers will be available on the site in Novosibirsk . Companies will also be able to use the cluster as a backup site to quickly restore IT systems in the event of accidents at the main location. For this purpose, customers can access virtual machine backup services based on Veeam Cloud Connect and VMware vCloud Availability disaster recovery (DRaaS). These solutions will minimize the simplicity of customer IT services and restore their availability in minutes.

"Thanks to the capabilities of the federal network" data centers -Rostelecom DPC, "the geography of clouds the DataLine has expanded significantly over the past year, covering new regional sites. Together with the established connectivity between sites, this will allow customers to quickly connect to our services throughout. to the country Companies will also be able to create geo-distributed solutions for backing up their systems and working next to their customers, "said Yuri Samoilov Rostelecom-Data Center First Deputy General Director, DataLine General Director.

The Rostelecom-Data Center in Novosibirsk was launched in early 2021. The capacity of the site is 304 posts, and the capacity is 3MW. In terms of reliability, the data center complies with the Tier III standard. The cluster also includes a new VMware NSX-T network virtualization solution, which will provide data center customers with a reliable data transfer environment.

Run ready-made virtual machine templates for Data Science

DataLine On April 27, 2021, the company announced the launch cloud of ready-made templates virtual machines for. Data Science Customers can now virtual data center quickly deploy the 18.04-based workplace in their analytics data own business. OS Ubuntu

The template already includes the necessary tools:

• Jupyter Notebook и JupyterLab; 
• conda package manager with Python 3
• libraries for machine learning, linear algebra and statistics: NumPy, scikit-learn, SciPy, CatBoost, LightGBM, XGBoost, TensorFlow, PyTorch; 
• control system; containers Docker 
• and other tools that facilitate the work of the analyst. 

Once the template is deployed, the analyst will have full access to the operating system and can add the necessary components.

The advantage of the workplace cloud is that the speed of learning models does not depend on the speed of the local, computer explained in the DataLine. The customer has the opportunity to choose the right amount, processors RAM, fast. disk storages The cloud runs on servers with 6254 processorsIntel Xeon v6 with a frequency of 3.1 GHz and support for AVX-512 instructions. To train complex models in private solutions, customers can get servers   with graphics accelerators for computing.

2020: Restart with Infinidat Storage

On January 14, 2020, the company DataLine announced that it had restarted the disaster-tolerant cloud on the base. SHD Infinidat

A disaster-tolerant cloud will keep business-critical applications running without the availability of one of the sites.

The disaster-tolerant cloud is based on a geographically distributed cluster virtualizations VMware based on two OST and NORD sites. The distance between the sites is 33 km. All cluster elements are duplicated on two sites. Synchronous replication is configured between the storage system so that the second site has an identical copy. virtual machine If one of the sites fails, the client will be able to switch to the standby without loss (data RPO = 0). At the same time, the service break will be no more than 3 minutes.

Customers of the catastrophic cloud of DataLine are already banks, retailers, e-commerce - all those business areas for which every minute of downtime entails significant financial and reputational losses.

The updated solution uses Infinidat InfiniBox F-series storage. The solution ensures high availability of storage, in the event of a failure of LUNs InfiniBox can independently switch to their copies on the backup site.

The InfiniBox has a convenient interface that simplifies initial setup and subsequent operation. The performance of each storage system, in addition to DataLine engineers, is monitored remotely by dedicated Infinidat specialists. This helps you get vendor assistance as quickly as possible in case of equipment problems.

2013: Start of a disaster-proof cloud of CloudLine Metrocluster

On October 1, 2013 , DataLine launched the service - the disaster-tolerant cloud CloudLine Metrocluster. According to the company, it provides "complete protection of business-critical applications and data from internal failures and external threats."

The solution runs on hardware NetApp with VMware vSphere software, and it is based on duplication of storage systems at two geographically separated sites - data centers DataLine Nord and Ost, located 33 km from each other. If one of the data centers is unavailable, virtual machines will be automatically started at the second site. The recovery speed of the virtual environment is between 2 minutes.

When one of the cluster nodes fails, application continuity must ensure synchronous data replication. An additional advantage of the new solution in DataLine is the possibility of load balancing at the data center network level.

Previously, a disaster-tolerant solution was available only within the framework of a private cloud, which was deployed for one client, explained TAdviser in the company. Now it is provided in the public cloud.

"The probability of any cataclysms is incomparably small, so this solution is provided as an option. Thus, our customers have a choice, "they add to the DataLine

According to the company's representatives , the SLA-agreement (service level agreement) of the new service provides for the technical features of a disaster-tolerant solution and changes will be made: in particular: the single downtime of the service will be reduced from 45 to 25 minutes, the total downtime per year (for technological work) will be reduced from 1.6 hours to 50 minutes.

The software developer for hosting and cloud services Parallels believes that the service is designed for large corporate customers who even minute downtime of the service can cause significant financial damage.

The company also notes that "there is a common misconception that the cloud service should be disaster-tolerant by default." Parallels representative Julia Yasnovskaya believes that this is not so: "if you need to ensure complete disaster tolerance, then you need to duplicate resources in different data centers, and also be able to manage it all."

"However, in any case, the reliability of the average cloud service is much higher than the average internal solution, since cloud services are provided from professional data centers that use a large number of reliability solutions that are practically unattainable for an ordinary enterprise user," she adds.

The company initially used the Double-Take Software system, which allows asynchronous replication of files of physical and virtual application servers, automates procedures for determining emergency situations and recovery, and is able to work with "weak" data channels. According to the DataLine, its high price significantly influenced the cost of relevant services for end customers and, as a result, was the main factor in low demand.

Therefore for CloudLine Metrocluster service products of Netapp, program obespechenieVMware vSphere, vCenter Site Recovery Manager and some other decisions of VMware were used. The service is based on the duplication of storage systems in two geographically spaced sites of the data center.

Each data center has a 30 percent reserve of computing resources relative to their peak load.