Between last august and today, you saw a lot of maintenance operations on Board Game Arena.
Almost all these maintenance operation concerned our "realtime server". This server is the one that allow you to see realtime updates on BGA. For example, when you play a move or create a table, the other players must see this immediately: this is the role of the realtime server.
During last summer, we found out that our existing realtime server was having difficulties to handle the growing traffic on BGA, causing red errors "TIMEOUT with notification service" you may have seen.
During the 3 last months, we performed a lot of upgrade on the realtime server. In particular, we configured a new server with 4 times more capacity than the old one, we did a lot of optimizations by rewriting a lot of code on this area.
The following graph is showing the average response time of one of our game server (in milliseconds). The yellow part is the time dedicated to the realtime server. As you can see, during the summer time, there were "peak" that was sometimes causing some "TIMEOUT with notification service". With all the optimizations, we manage to reduce this time to almost nothing, and to lower by 40% the average response time of our server.
We know that all these maintenance operations are painful. They are especially painful when we change something on this server, because all the games are using it so we must interrupt the whole service when working on this one. However, we'd like to share with you this graph to show you that we are not doing this for nothing, and that there is a benefit for you after all of this.
Almost all these maintenance operation concerned our "realtime server". This server is the one that allow you to see realtime updates on BGA. For example, when you play a move or create a table, the other players must see this immediately: this is the role of the realtime server.
During last summer, we found out that our existing realtime server was having difficulties to handle the growing traffic on BGA, causing red errors "TIMEOUT with notification service" you may have seen.
During the 3 last months, we performed a lot of upgrade on the realtime server. In particular, we configured a new server with 4 times more capacity than the old one, we did a lot of optimizations by rewriting a lot of code on this area.
The following graph is showing the average response time of one of our game server (in milliseconds). The yellow part is the time dedicated to the realtime server. As you can see, during the summer time, there were "peak" that was sometimes causing some "TIMEOUT with notification service". With all the optimizations, we manage to reduce this time to almost nothing, and to lower by 40% the average response time of our server.
We know that all these maintenance operations are painful. They are especially painful when we change something on this server, because all the games are using it so we must interrupt the whole service when working on this one. However, we'd like to share with you this graph to show you that we are not doing this for nothing, and that there is a benefit for you after all of this.