Yandex database performance improving

Yandex-fintech-antifraud1. High performance2. Cluster reliability3. Designed common graph-based system for manager’s-configured-logic execution.4. Cluster information security5. Personal data security6. Personal data gluing through user features7. Ask external services for additional data8. Dynamic history data accumulation9. lua/js/custom languages usage for manager scripts

Yandex-logistic-platform1. High performance2. cluster reliability3. Interaction with logistic operators4. Integration platform for transfer optimization5. a lot of business processes with logistic (there are really a lot of processes!!!))))Yandex-logistic-dispatcher1. Algorithms of dispatch (very fast and productive). courier path duration reduced.2. claster reliability3. Interaction other services in taxi cluster4. Prepare new lead

Yandex-Drive (carsharing).1. Money control in system2. Algorithms on graph (cluster)3. Robots cluster (control some background processes for service-workers and users)4. Interaction with real cars5. Billing processes6. System of integration all components7. Tag-based logic for control cars8. User permissions control system9. Model of system

Graph-Algorithms cluster1. just routing2. matrices3. waves for detect optimal positions for routing4. graph scaled for cluster5. graph routing quality6. high performance7 reliability

a. Data delivering 1. Reliable guarantee delivering data for index makers. 2. High performance storage for non delivered documents collection. 3. Data receiving with client-api for custom services. 4. Data receiving from big data storages in async mode by subscribing.b. Searching 1. Guarantee collect data for search requests from many replicas and shards of index. 2. Consistency options. 3. Full web searching. 4. Parametric search for custom services.… Show more a. Data delivering 1. Reliable guarantee delivering data for index makers. 2. High performance storage for non delivered documents collection. 3. Data receiving with client-api for custom services. 4. Data receiving from big data storages in async mode by subscribing.b. Searching 1. Guarantee collect data for search requests from many replicas and shards of index. 2. Consistency options. 3. Full web searching. 4. Parametric search for custom services. 5. Key-value storage on base searching. 6. Suggest for search as ordinary system feature. 7. Common user api for all use-cases.c. Index construction 1. Index construction with optional features selection for different cases (simple-search/web-search/suggest/kv-storage/etc...). 2. Multi-threading architecture for index construction speed up. 3. Disk utilization optimizing. 4. Index recovering for different use-cases in case emergency stop. 5. Additional features implements as plugins for main system. 6. Processing errors in case impossible index recovering.d. Automatic control of the cluster 1. Start new service in 3 clicks on web-interface. 2. Service automatic configuration by performance and functional require-ments. 3. Automatic servers allocation and deploy configuration on start. 4. Control for data status on servers. In case data corruption or broke server, recovering data occurs automatically 5. Configuration optimization by network topology.e. Testing 1. Cluster local model. 2. Cluster distributed model. 3. Coverage about 89%. 4. Reliable new features implementation. 5. Deploy binaries system check testing before deploy. 6. Almost all production problems have models in testig Show less