How to improve site performance by 8 times: our experience in integrating AWS services

The software development outsourcing company AgiliWay together with the distributor Softprom by ERC delivered and implemented AWS services in ROI4CIO and increased the site's performance 8 times.

Project Dates

09.28.2017 - 10.10.2017: selection of the optimal server configuration, selection of optimal services on AWS and full transition to AWS, including the domain name.

Problem

The service used on the site prior to AWS implementation did not withstand work with "heavy" functionality, in particular, the need to process a large amount of information. As a result, with a certain number of users, the CPU and RAM indicators of the server rose to 100% and the site crashed.

Result

Thanks to the use of the server and AWS services, the load on the site and the database has decreased by 2 times, the site can withstand several times more users at the same time. The site's performance was increased by 8 times, while the cost of ownership has increased only 2 times.
')
ROI - 800%

Problem

“The team faced a problem with the lack of site performance. During testing, while connecting a certain number of users, the response delay was more than 10 seconds. When the load increased, the base did not withstand, the service “refused” to work. ”

Oleg Pitsyk, IT Systems Architect ROI4CIO, Agiliway.

Decision

Before making a decision on the implementation of AWS services for the ROI4CIO project, a possible placement on another cloud resource was tested. AWS, Vultr, goDaddy, Linode and others were analyzed. Comparative test results showed a clear advantage of AWS services.

ROI4CIO Pre-AWS Architecture


As a result of interaction, Softprom by ERC together with Agiliway developed the optimal solution to improve the performance of the site. Representatives of Agiliway, as a developer and architect of systems ROI4CIO, were engaged in setting up an application server and database. Softprom by ERC acted as a supplier of AWS platforms and as a consultant on the functional use of AWS services. In addition, Softprom by ERC employees were engaged in setting up monitoring, notifications and automation of Lambda.

The site domain is hosted on Amazon Route 53. This is a highly available and scalable cloud-based web service of the domain name system (DNS).

To provide comprehensive protection against all known infrastructural attacks (level 3 and 4), the anti-DDoS AWS Shield system was used.

To optimize the workloads that require large computational power, we used the EC2 virtual server type C4, as well as configured auto scaling resources and balancing the incoming load on the server. At times of peak site load, computing resources are automatically scaled, which allows you to withstand almost any load.

In order to improve reliability and fault tolerance, as well as reduce administrative costs, a database service Amazon Relational Database Service was chosen to host the database, which performs the functions of allocating hardware, setting up a database, installing patches, and backing up.

To increase the speed of the site, ElastiCashe was also used - a web service that simplifies deployment and scaling in the cloud storage or memory cache.

Using AWS CloudWatch, advanced monitoring of the application and database was configured. Creating various rules, the administrator immediately receives an SMS notification to the mobile and to the e-mail in case of any unforeseen situation. To implement the function of sending messages to CloudWatch, the simple notification service (AWS SNS) was integrated. In addition to sending messages, under certain circumstances, triggers are triggered, which activate the automation functions implemented using the AWS Lambda serverless computing service.

Backups of the database and application server are automatically created according to the schedule and saved to the AWS S3 cloud storage.

Updated architecture

“Basically, all efforts were aimed at increasing productivity. But also a lot of attention paid to fault tolerance and backup. By numbers, you can probably allocate more bandwidth disk storage base (IOPS). Used a disk with a high bandwidth Provisioned IOPS. Up to 40,000 IOPS can be allocated to each DB Instance. "

Vlad Gavrilenko, IT Director, Softprom by ERC.

“We moved the server very quickly, in just 2 days. Then we were setting up services. It took about a week. ”

Vlad Gavrilenko, IT Director, Softprom by ERC.

Result

Thanks to the introduction of AWS services, the team provided an eight-fold increase in resource productivity and the simultaneous automation of time-consuming administrative tasks. At the same time, the cost of ownership has only doubled. After correction of the request code and the introduction of all services, the site maintains a continuous and stable performance even with high loads.