How to combine backups of two retailers on SAP in 12 hours

This article will tell you about a large-scale SAP implementation project in our company. After the merger of M.Video and Eldorado companies, the technical departments were faced with a nontrivial task - to transfer business processes into a single backend based on SAP.

Before the start, we had a duplicated IT infrastructure of two chain stores, consisting of 955 retail outlets, 30,000 employees, and three hundred thousand checks per day.

Now that everything is up and running successfully, we want to share a story - how we managed to make this project.
')
In this publication (the first of two things that the devil does not joke, maybe even three) we will present you some data on the work done, which you can learn more about at the SAP ME meetup in Moscow.



Half a year of design, half a year of writing code, half a year of optimization and testing. And 12 hours to launch a common system in 1,000 stores throughout Russia (from Vladivostok to Kaliningrad).

It may sound unrealistic, but we did it! Details under the cut.

In the process of uniting the companies M.Video and Eldorado, we were faced with the task of optimizing costs and reducing the business processes of two different companies to a single backend.

Perhaps this can be called luck or coincidence - both retailers used SAP systems to organize processes. We had to deal only with optimization, and not with complete restructuring of the internal systems of the Eldorado network.

Functionally, the task was divided into three (actually four) stages:

1. Designing "on paper" and coordination by our business analysts and SAP consultants of new processes (as well as modernization of old ones) within the already operating systems.
   
   After analyzing a number of indicators of the already working backend of the two companies, the M.Video backend was taken as the basis for developing a unified system. One of the main criteria by which the choice was made - the efficiency of the company as a whole, the greater revenue and profit at lower costs for a business transaction.
   
   At the stage of analysis and design, it took about six months, billions of nerve cells of department heads and technical specialists, and very, very many liters of coffee were drunk.
2. Implementation in code . We give a few figures on the results of the project:
   
   • 2,000 routes per day, planned through a logistics module.
   • 38,000 users front and backend.
   • 270 000 goods in warehouses of the united enterprise.
   
   About 300,000 checks processed by the system per day, which are subsequently stored for up to five years in order to provide customers with a guarantee, as well as for marketing research.
   
   Calculation of salaries, advances and bonuses for 30,000 employees every month.
   
   The project involved a team of 300 technical specialists who worked for ten months. By simple arithmetic calculations we get two figures, clearly showing the scale of the work done: 90,000 man / days and ... 720,000 hours of work .
   
   
   
   Further, the optimization of individual subroutines of the SAP modules, about a hundred of subroutines, was accelerated five to six times by optimizing the code and queries to the database.
   
   For individual cases, we managed to reduce the program execution time from six hours to ten minutes by optimizing queries to the DBMS
3. The third stage is probably the most difficult - testing . It consisted of several cycles. To hold them, we assembled a team of 200 employees, they were involved in a functional, integration and regression test.
   
   We will describe stress tests in a separate paragraph, they consisted of 15 cycles for each of the SAP modules: ERP, POS, DM, PI.
   
   According to the results of each test, optimization was made of the code and parameters of the DBMS, as well as the database indexes (they work for us on SAP HANA, part on Oracle).
   
   After all the load tests, about 20% percent was added to the computational computing capacity, and a reserve of approximately the same (20%) volume was also formed.
   In addition, after carrying out the above cycles, an analysis of 100 of the most resource-intensive programs began, according to the results of which we refactored the code and accelerated their work on average five times (which once again confirms the importance of refactoring and code optimization).
   
   The last of the tests was "cut over". A separate test zone was created for it, which copied our productive data center. We did “Cut over” twice, each time it took about two weeks, in which we measured the speed of operations such as transferring program settings from the test zone to the productive one, loading open positions on stocks of goods and periods of unavailability of operations.
4. And the fourth stage is the immediate launch after passing the tests. The task was, frankly, difficult: in 12 hours to switch about 955 stores across the country, and not to stop sales.

On the night of February 24-25, a team of the ten best specialists of our company took over the “watch” in the data center, and the transition magic began. We will tell about it in detail at our meetup, and after the second article we will devote the technical details of our SAP magic.

Results.

So, the result of the work was an increase in such indicators as:

• The load on the backend has approximately doubled.
• The number of checks per day increased by 50% from 200 thousand to 300 thousand.
• The number of frontend users increased from 10 thousand to 20 thousand.
• The module for calculating salaries has increased the number of employees from 15 thousand to 30 thousand people.

We will talk about all the technical details on our SAP mitap in Moscow, which will be held on June 6 at the M.Video-Eldorado office. Experts will share their implementation experience. According to the results of the meeting, young specialists will be able to get a paid internship in the company with the prospect of further employment.

Learn more and register at this link.