A telecom operator of MegaFon’s level manages an incredible workload.
MegaFon already has 80 million users
The billing system of MegaFon processes 2 billion transactions
Of data change actively, 8 PB of data change a little more slowly. All this data is live, not historical; it’s used in a single billing process
5 000 servers in 14 data centers ensure service operation and payment
In 2016, the external audit of the IT landscape in MegaFon has shown that it is necessary to improve fault tolerance and availability of basic structures: billing, ERP, operations and business support systems.
MegaFon has decided to move to a microservice architecture and rebuild the IT landscape for a three-layer model. This move solved the load on the underlying systems and accelerated time to market.
The concept assumes that basic structures are located on the bottom layer, which practically do not change. This includes the core of billing, communication systems, ERP.
In the middle are microservices, which are updated every few weeks. This layer contains caches, data marts, marketing systems, etc.
The top layer is the most mobile. The client interfaces are presented here, so it takes several days from setting a business task to updating.
It has been decided to take off the load from the core to the middle-level microservices using a lightweight and reliable storage. MegaFon has formed the following requirements:
Tarantool does not replace Oracle but complements it
Tarantool-based solutions support services for 80,000,000 subscribers in 10 time zones
25+ services within MegaFon use Tarantool with regard to geographic redundancy and receive updates without downtime
Tarantool experts provide turnkey services and help create an internal competence center in MegaFon
The developers of MegaFon create Tarantool-based solutions on their own: the internal development team has grown from 4 people to 80+ in two years
Tell us about the tasks of your project,
and we will compose a solution for you on Tarantool
Tarantool experts will
contact you shortly