Performance Optimization Practices for iQIYI International Mobile App

When targeting a global audience, network performance is critical. iQIYI International observed that a 10‑second homepage load would cause massive user loss, so they conducted a series of optimizations to improve overseas user experience.

Core optimization ideas : avoid unnecessary network requests, aim for 0‑RTT connections, and keep request payloads as small as possible.

Network research : Sampling request chains in early app versions revealed regional issues such as ISP hijacking, complex mobile networks in Southeast Asia, a significant share of low‑end Android devices, and high client‑to‑server latency. Figures show latency from overseas users to data centers and mobile network conditions in major regions.

Identified problems :

Severe ISP/DNS/HTTP hijacking.

Complex mobile networks, especially in Southeast Asia.

Notable proportion of low‑end Android devices.

High end‑to‑server latency.

Technical work focused on solving these issues before scaling operations.

Improvement roadmap :

01 Basic link optimization

DNS optimization : Switch to HTTPDNS, which improved initial connection efficiency by 17% and mitigated domain hijacking and non‑local DNS allocation. Pre‑warming of DNS entries is also possible.

MTU issues : Mismatched MTU between client and server (e.g., AWS jumbo frames of 9001 bytes vs. default 1500 bytes) caused packet loss. VPN tunnels and networks that block ICMP prevent automatic MTU discovery.

TCP layer optimization :

Increase congestion window (CongWin) based on bandwidth‑delay product; default CentOS values are far below the theoretical maximum.

Enable TCP Fast Open (TFO) to reduce connection re‑establishment overhead, yielding ~12% HTTP request efficiency gain in overseas scenarios.

02 Application‑layer optimization

HTTP improvements : HTTP/1.1 keep‑alive is less effective for mobile apps with scattered requests; upgrade to HTTP/2.0 for framing, multiplexing, and header compression, eliminating head‑of‑line blocking.

SSL/TLS enhancements : Adopt TLS 1.3 with PSK and early data for 0‑RTT, or at minimum use ECDHE cipher suites to reduce handshake RTTs from 2 to 1.

Upgrade to HTTP/2.0 : Benefits include frame‑based transmission, multiplexed streams, and HPACK header compression, which prevents oversized cookies.

03 Transport content optimization

Brotli compression : Provides ~17% better compression than gzip, reducing average response size from 30 KB to 18 KB.

Protobuf over JSON : Parsing speed improves 4‑5×; however, varint compression only helps for numbers <128, so gzip remains necessary for large payloads.

WebP images : Switching to WebP with 85% quality cuts image size by ~45% compared to JPEG/PNG, noticeably speeding up homepage image loading.

04 Business‑layer improvements

Reduce unnecessary requests : Cache static content (navigation, channels) and only fetch updates when timestamps differ.

Network‑aware strategies : Serve lower‑resolution video (360p) on non‑WiFi and provide dual‑quality image URLs (high‑quality vs. low‑quality for 4G).

Additional tweaks : Add request retries, adjust HTTP timeout, and enable fallback protocols (QUIC or KCP) for failed requests, improving success rate by 45%.

Overall, these optimizations significantly lowered latency, reduced client failure rates, and boosted video playback success, helping iQIYI International achieve top‑rank positions in several Southeast Asian markets.