What is Write-Ahead Log (WAL) Compression in PostgreSQL? In which cases does enabling WAL file compression reduce the load on the disk subsystem and network, and in which cases does it lead to performance degradation due to excessive CPU core utilization?
How is the Event Loop Policy mechanism structured in asyncio? In which scenarios does a backend developer need to forcibly change the default event loop policy to a custom one, for example using uvloop based on libuv, and how is the performance gain achieved?
What is an expression index (Expression Index / Functional Index) in PostgreSQL? What are the overheads on write operations (INSERT/UPDATE) it entails and why might the planner refuse to use it if the data types in the query expression do not match?
What is deep copying of objects (copy.deepcopy) in Python from a performance perspective? How does this mechanism bypass cyclic references, why does it work critically slowly on large object graphs, and what alternative patterns replace it in high-load systems?
How does the Sidecar pattern work in Kubernetes, for example with Istio or Envoy? How does this pattern allow offloading traffic encryption (mTLS), request tracing, and retries from the application?
What are Cache Penetration, Cache Avalanche, and Cache Warmup? How to design a caching layer for an online store so that during a night sale, when millions of users simultaneously request new products, the database does not fail due to cache misses?
What are the differences between Push-based and Pull-based architectures in designing data streaming systems like RabbitMQ and Kafka? How do these approaches affect memory utilization on the broker side and the issue of consumer backpressure?
Describe the two-phase commit (2PC) protocol for distributed transactions involving two independent relational databases. What happens to the system if the transaction coordinator crashes exactly between the voting phase (Prepare) and the commit phase?
Describe the mechanics of the built-in compile() function. How to convert a dynamically generated string of Python code into a code object, execute it safely in an isolated context using exec/eval, and prevent remote code execution vulnerabilities (RCE)?
Describe the trade-offs between WebSockets, SSE (Server-Sent Events), and gRPC Streaming protocols. Which protocol would you choose for: a) live chat, b) one-way stock ticker feed, c) inter-service communication in microservices architecture, and why?
How are C/C++ extensions for Python (C Extensions) built under the hood? Why do extensions written using the traditional Python C API slow down the transition to No-GIL Python, and how do modern alternatives like HPy or Cython solve this problem?
Why does the standard Python function os.cpu_count() often return the total number of cores of the physical server inside a Docker container running in a Kubernetes pod with CPU limit: 2? What critical problems with thread pool size in Gunicorn or Celery does this cause, and how to fix it?
How does the Consistent Hashing algorithm work? Why is it needed in designing a distributed cache cluster, for example with 20 Redis servers, and how does it minimize the number of keys that need to be moved when adding or removing a node?
How do Linux Cgroups (Control Groups) and Namespaces create isolation within a Docker container? What happens to a Python process when it exceeds the resource.limits.cpu in Kubernetes, and why does exceeding limits.memory lead to OOMKilled, while exceeding CPU only causes throttling?
