What does Digg work on
The main architect of Digg.com Joe Stump (Joe Stump) told in a corporate blog about how the largest news aggregator is currently located and works at 142 place in attendance in the Alexa.com ranking.
“If you ask Ron, our chief system engineer, how many servers we have now,” says Joe, “he will most likely say honestly that he doesn’t know.” It is absolutely true that there are 6 servers of a graph database for a recommendation mechanism - a complex mathematical algorithm that tells readers news that may seem interesting to them. Graphic DBMS was developed by Digg specialists specifically for this purpose, since the usual relational ones were not suitable for it. There are 6 file servers for images running MogileFS. In total, the number of cars in the tens. But the number of servers is not the most interesting thing to learn about the Digg architecture.
The first thing that a user request meets on its way is load balancing servers. They perform several tasks: they monitor the cluster of applications and input / output its components as needed, distribute requests to the cluster servers and cache JavaScript, CSS, and images. Joe advises Linux Virtual Server and Squid as a free counterpart to these servers.
')
Application servers are split into several daemons, including Apache + PHP, Memcached, and Gearman. The latter act as "conductors", distributing requests from applications to the most accessible file and database servers at the moment.
Interestingly, MogileFS, Memcached, and Gearman were developed by Danga Interactive, which created them for Livejournal, even when it belonged to Brad Fitzpatrick. Six Apart, which bought it, renamed it Livejournal Inc. What happened to the company is well known to all.
The database is distributed between 4 master servers, which are assigned to a heap of slaves. On master'ah write operations occur, on slave'ah - read. All of them work under MySQL.
“If you ask Ron, our chief system engineer, how many servers we have now,” says Joe, “he will most likely say honestly that he doesn’t know.” It is absolutely true that there are 6 servers of a graph database for a recommendation mechanism - a complex mathematical algorithm that tells readers news that may seem interesting to them. Graphic DBMS was developed by Digg specialists specifically for this purpose, since the usual relational ones were not suitable for it. There are 6 file servers for images running MogileFS. In total, the number of cars in the tens. But the number of servers is not the most interesting thing to learn about the Digg architecture.
The first thing that a user request meets on its way is load balancing servers. They perform several tasks: they monitor the cluster of applications and input / output its components as needed, distribute requests to the cluster servers and cache JavaScript, CSS, and images. Joe advises Linux Virtual Server and Squid as a free counterpart to these servers.
')
Application servers are split into several daemons, including Apache + PHP, Memcached, and Gearman. The latter act as "conductors", distributing requests from applications to the most accessible file and database servers at the moment.
Interestingly, MogileFS, Memcached, and Gearman were developed by Danga Interactive, which created them for Livejournal, even when it belonged to Brad Fitzpatrick. Six Apart, which bought it, renamed it Livejournal Inc. What happened to the company is well known to all.
The database is distributed between 4 master servers, which are assigned to a heap of slaves. On master'ah write operations occur, on slave'ah - read. All of them work under MySQL.
Source: https://habr.com/ru/post/39685/
All Articles