Why CDNs Provide Little Acceleration for Mobile Clients

Google performance engineer and author Ilya Grigorik recently published a blog post titled “Why CDN Has No Effect on Mobile Client Acceleration,” describing the peculiarities of mobile (wireless) networks and proposing a concept for a mobile‑friendly CDN.

Grigorik criticizes current CDNs for their poor acceleration on mobile devices; monitoring data shows that traditional CDN optimizations have little impact, prompting a call for a CDN architecture that better supports mobile networks.

He identifies two common misconceptions: (1) traditional CDNs provide roughly the same absolute performance boost for mobile clients as for broadband, and (2) the issue is not the existence of a “wireless CDN” but rather the carrier network itself.

He provides reference latency data to analyze the main components of wireless network delay:

Client located on the West Coast, server on the East Coast.

Coast‑to‑coast network latency: 50 ms.

Server response latency: 50 ms.

Shared client last‑mile latency: fiber ≈18 ms, cable ≈26 ms, DSL ≈44 ms.

Wireless client last‑mile latency: 4G ≈50 ms, 3G ≈200 ms.

Note: “Last‑mile” refers to the segment from the carrier’s exchange equipment to the end‑user’s device.

The following diagram illustrates the user access flow and latency when using a CDN:

Using a CDN for Content Delivery Acceleration

CDN acceleration requires deploying cache servers at many peering points worldwide, placing content as close to users as possible. Ideally, a CDN server can immediately locate the client’s ISP, and the last‑mile latency consists of the client‑to‑ISP segment plus the CDN server’s response time.

CDNs reduce propagation latency.

When static resources are cached, CDNs also reduce server response time.

Continuing the earlier example, assume the CDN has optimized the network so that the coast‑to‑coast latency drops from 50 ms to 5 ms and the client‑to‑CDN latency (when the request is a cache miss) is 5 ms. For a fiber client, the total time becomes 18 ms (last‑mile) + 5 ms + 5 ms + 5 ms + 18 ms = 51 ms, a 365 % improvement over the original 186 ms.

The table below compares performance metrics with and without CDN acceleration for various connection types:

Last‑mile

Coast‑to‑Coast (low)

Server Response

Total (ms)

Improvement

Fiber

18

50

50

186

Cable

26

50

50

202

DSL

44

50

50

238

4G

50

50

50

250

3G

200

50

50

550

CDN + Fiber

18

5

5

51

-135 ms (365 %)

CDN + Cable

26

5

5

67

-135 ms (301 %)

CDN + DSL

44

5

5

103

-135 ms (231 %)

CDN + 4G

50

5

5

115

-135 ms (217 %)

CDN + 3G

200

5

5

415

-135 ms (133 %)

Repeating the same calculations for each connection reveals an unfortunate trend:

The higher the last‑mile latency, the less effective the CDN appears.

Since CDN servers are usually placed outside the ISP network, node selection becomes crucial.

CDNs can still provide some improvement to last‑mile latency.

CDNs help reduce propagation and server response times, but when measuring before‑and‑after performance, the gains for mobile clients are modest; for example, 3G users typically see only about a 33 % improvement.

Operational and Business Maintenance Costs at Edge Nodes

A clear strategy is to move cache servers closer to customers—ideally inside the carrier’s network rather than in external peering points. Deploying nodes inside carriers is theoretically possible, and many carriers already operate their own caches, but practical challenges arise:

Peering points are limited; deploying mobile servers inside each carrier would require separate settlements, so most servers remain in shared data centers.

Placing servers near customers (e.g., at antenna towers) would demand massive hardware deployments, leading to operational nightmares, security concerns, and the need for third‑party TLS termination.

Carriers have long tried to offer CDN services, yet signing individual agreements with each carrier is rarely attractive to website owners.

Recent news reports that Verizon acquired EdgeCast suggest that such integration could benefit Verizon’s customers.

Beyond operational costs, CDNs provide no special optimizations for mobile clients. The core issue is the poor last‑mile latency of mobile carriers, which must be addressed by improving the network itself and encouraging competition among carriers, rather than merely moving caches closer to users.

In China, the carrier environment is even more complex, with many large and small operators. Smaller ISPs are expanding into secondary cities, and various regional networks add to the fragmentation.

Because inter‑carrier settlement fees exist, carriers strive to keep traffic within their own networks, leading to “traffic hijacking” practices that become increasingly sophisticated.

The domestic CDN market is highly competitive, with major players (e.g., Wangsu, Blue Cloud, Kuaiwang, DiLian) partnering with carriers, launching mobile acceleration solutions, and integrating CDN with cloud services, intensifying the competition.

This complexity makes it harder to resolve user access issues; some argue that only broader internet reforms will improve the situation, but closer collaboration between carriers and content providers can steadily enhance user experience.