25 billion requests per hour: ServiceNow database
On February 26, the MariaDB M | 18 conference was held , in which 330 IT companies participated. A representative of ServiceNow spoke at the conference. He shared the case management "85 thousand databases around the world." The average number of requests per hour is 25 billion. As in ServiceNow work with such loads, we will tell under the cut.
/ Flickr / tethys. / CC
ServiceNow spokesperson Tim Yim spoke about using MariaDB. According to Yima, the company chose MariaDB because this database is stable and highly accessible. At ServiceNow, we focus on 99.999% availability of systems, because this indicator is critical for customers such as hospitals, factories, banks and power plants.
')
MariaDB has already appreciated the benefits of using DBS Bank in Singapore by abandoning DB2 and Oracle Enterprise. Over the past two years, the bank has deployed more than 700 copies of MariaDB DB, which employ 54% of business critical services, conducting daily transactions of 70–80 billion dollars. The company increased application performance seven times, and the tools for testing them were able to handle ten times the load. At the same time, automated application release systems have become twice as efficient.
The company says they do not use multi-tenant and single-tenant architecture. Instead, ServiceNow came up with its option - multi-instance. The multi-instance deployment model allows each client to get its own database and client application.
Yim explains : “Usually, companies deploy a multi-tenant architecture when all clients share one large database. Imagine that there are a thousand customers. Then the failure of one of them will affect 999 others: in the end, instead of a single idle, you get a thousand. ”
In the case of multi-instance, each client database instance runs on bare metal for better performance. At the same time, hardware resources are divided between the levels of applications and databases, and the processes are “containerized”. Yim commented : “We are duplicating every configuration, every node of client applications, and every database. Every night we create backup copies of all databases. ”
This approach allows you to increase computing power with "surgical precision." At any given time, you can scale the infrastructure and troubleshoot a single client.
/ Flickr / david goehring / cc
Yim said that in order for this solution to work effectively, "you need to automate everything." In ServiceNow did so. All assets are loaded into CMDB so that they can be tracked in one place, and then workflows are configured and automated: high availability systems, failover systems, etc. In the latest releases of the ServiceNow platform, developers added the ability to automate natural language, allowing non-specialists ”to automate the necessary processes in their department.
Managing this amount of DB leads to a number of problems. Tim Yim explains: “The dynamism of ServiceNow allows our clients to scale services, but because of this, query patterns change. Therefore, such a platform is very difficult to manage. "
MariaDB employees help to solve all ServiceNow problems: they provide trainings for database administrators, system administrators and infrastructure reliability specialists (site reliability engineer), and also help ServiceNow to understand technical issues.
The companies plan to continue collaboration and implement features such as DDL to describe the database structure and more secure data encryption in the cloud. In addition, a spokesman for ServiceNow said he was not against the introduction of machine learning-based systems: “I think that any DBA would be interested in working with a smarter database that can help in load distribution.” And the developers of MariaDB are already working on the implementation of monitoring tools that will improve the performance and scalability of the system.
A few more posts from the Guild IT corporate blog:
/ Flickr / tethys. / CC
Why use MariaDB in ServiceNow
ServiceNow spokesperson Tim Yim spoke about using MariaDB. According to Yima, the company chose MariaDB because this database is stable and highly accessible. At ServiceNow, we focus on 99.999% availability of systems, because this indicator is critical for customers such as hospitals, factories, banks and power plants.
')
MariaDB has already appreciated the benefits of using DBS Bank in Singapore by abandoning DB2 and Oracle Enterprise. Over the past two years, the bank has deployed more than 700 copies of MariaDB DB, which employ 54% of business critical services, conducting daily transactions of 70–80 billion dollars. The company increased application performance seven times, and the tools for testing them were able to handle ten times the load. At the same time, automated application release systems have become twice as efficient.
What is a multi-instance architecture?
The company says they do not use multi-tenant and single-tenant architecture. Instead, ServiceNow came up with its option - multi-instance. The multi-instance deployment model allows each client to get its own database and client application.
Yim explains : “Usually, companies deploy a multi-tenant architecture when all clients share one large database. Imagine that there are a thousand customers. Then the failure of one of them will affect 999 others: in the end, instead of a single idle, you get a thousand. ”
In the case of multi-instance, each client database instance runs on bare metal for better performance. At the same time, hardware resources are divided between the levels of applications and databases, and the processes are “containerized”. Yim commented : “We are duplicating every configuration, every node of client applications, and every database. Every night we create backup copies of all databases. ”
This approach allows you to increase computing power with "surgical precision." At any given time, you can scale the infrastructure and troubleshoot a single client.
/ Flickr / david goehring / cc
Yim said that in order for this solution to work effectively, "you need to automate everything." In ServiceNow did so. All assets are loaded into CMDB so that they can be tracked in one place, and then workflows are configured and automated: high availability systems, failover systems, etc. In the latest releases of the ServiceNow platform, developers added the ability to automate natural language, allowing non-specialists ”to automate the necessary processes in their department.
"Hidden" difficulties and future plans
Managing this amount of DB leads to a number of problems. Tim Yim explains: “The dynamism of ServiceNow allows our clients to scale services, but because of this, query patterns change. Therefore, such a platform is very difficult to manage. "
MariaDB employees help to solve all ServiceNow problems: they provide trainings for database administrators, system administrators and infrastructure reliability specialists (site reliability engineer), and also help ServiceNow to understand technical issues.
The companies plan to continue collaboration and implement features such as DDL to describe the database structure and more secure data encryption in the cloud. In addition, a spokesman for ServiceNow said he was not against the introduction of machine learning-based systems: “I think that any DBA would be interested in working with a smarter database that can help in load distribution.” And the developers of MariaDB are already working on the implementation of monitoring tools that will improve the performance and scalability of the system.
A few more posts from the Guild IT corporate blog:
Source: https://habr.com/ru/post/350424/
All Articles