From the early fall, we are working on the reworking of our tracker database architecture. We succeeded to improve in redirects speed and reports building.
Currently, situation is as following
1. The response time on any load is about 30 ms! We couldn’t emulate load with hundreds of millions of clicks in a day. But I don’t think there are that many people who can buy this much traffic. Theoretically, the tracker can handle as much.
Algorithmically it is not that important how much of input traffic is sent. All current tests were done at volumes approximately 30 million in a day for difficult campaigns with lots of paths, landing pages, offers, and rules.
2.Unique system of stats aggregation was implemented. The tracker develops all necessary report data on the clicks stage recording. Thanks to which all base reports by campaigns, offers, landings pages, etc. are generated instantly (tens of ms) in any timeframes. I can’t disclose all the details, but this is really cool.
3.Over the past week we improved the report generation system and transferred successes with main reports to reports inside of a campaign. We managed to receive almost all reports at any depth instantly. We still experience some issues with tokens and referrals, but we are currently solving it and incorporating data into the interface. It looks amazing right now. Besides, the tracker can output data either per page or all together for any term. Generation speed of any report doesn’t depend on clicks quantity in any given campaign or the quantity in the database in general. So, you can instantly navigate through the pages of a report, minimize and maximize groupings, download CSV. Reports that couldn’t be generated in a wink, also work very fast, because the majority of necessary data is stored ready-made.
Next step is to learn how to generate different reports by offers, landing pages, sources, etc. For example, if you want to check report by all publishers for all time.
As for functionality, the tracker is currently at the same level of other trackers. You have all you need to work, but without superfluous features. The only thing I’d like to add in beta is bots analysis, which will provide the percentage of mobile bots on for publisher. It can analyze very precisely. A more accurate analysis is hardly possible.
In general lots of people offer their help. Thank you for not being indifferent to our work.
What started as “database optimisation” ended as a full reworking of tracker core, clicks recording algorithms and all stats generation. Currently, we are working on interface improvement, site design and billing.