Global Load Balancing Systems And Traffic Control Power

Introduction

Global load balancing systems are technological systems that distribute internet traffic, application requests, data flows, and computing workloads across multiple servers, data centres, cloud regions, content-delivery networks, and network operators. They are increasingly important in cloud computing, content delivery, search, e-commerce, financial infrastructure, streaming, online advertising, and digital platforms.

From a competition-law perspective, global load balancing can create a distinctive form of market power: traffic-control power. A company that controls the algorithmic layer determining where users' requests go, which server or network receives traffic, which infrastructure is preferred, and which competing infrastructure is bypassed may influence competition even without owning every underlying server or network.

The central competition question is therefore:

Can control over the routing and allocation of digital traffic constitute a source of market power or an essential competitive bottleneck?

Potential concerns arise under abuse of dominance, refusal to deal, discriminatory access, tying, self-preferencing, foreclosure, interoperability, essential-facilities principles, vertical restraints, and merger control.

1. Meaning of Global Load Balancing

Traditional load balancing distributes requests among servers within a relatively limited infrastructure.

Global load balancing (GLB) operates at a much larger geographical and infrastructural level. It can determine whether a request is routed to:

• a particular geographic region;
• a particular cloud provider;
• a particular data centre;
• a CDN node;
• a particular ISP;
• a particular application server;
• a particular API endpoint;
• a particular payment or identity service;
• a particular network path.

A simplified model is:

User → DNS/Anycast/Traffic Manager → Routing Algorithm → Data Centre/CDN/Cloud → Application

The routing decision can depend upon:

1. latency;
2. congestion;
3. geographic location;
4. server availability;
5. contractual preferences;
6. price;
7. network performance;
8. reliability;
9. security;
10. capacity;
11. business relationships; and
12. algorithmic optimisation.

Thus, the entity controlling the routing layer may become a traffic intermediary between users and competing infrastructure providers.

2. What Is “Traffic Control Power”?

Traffic-control power refers to the ability of an infrastructure operator or platform to materially influence the destination, quantity, timing, quality, or visibility of digital traffic.

It can be understood through five dimensions.

A. Destination control

The operator determines which infrastructure receives a request.

B. Volume control

It can increase or decrease traffic directed toward particular providers.

C. Quality control

It may influence latency, availability, bandwidth, or reliability.

D. Priority control

Some traffic may receive preferential routing.

E. Information control

The routing system generates valuable information about:

• user demand;
• geographic demand;
• congestion;
• infrastructure performance;
• competitor utilisation;
• peak periods;
• switching behaviour.

This information itself can become a competitive asset.

3. Why Global Load Balancing Can Create Competition Problems

A. Bottleneck control

Suppose a dominant cloud or CDN provider operates the principal traffic-management layer used by thousands of websites.

Even if competing clouds remain technically available, the dominant provider could make them commercially less attractive by:

• assigning them fewer requests;
• increasing routing latency;
• imposing technical restrictions;
• limiting interoperability;
• requiring use of proprietary APIs;
• charging discriminatory routing fees.

The bottleneck is therefore not necessarily the physical server.

It may be the decision-making layer that controls traffic reaching the server.

4. Traffic Control as a Digital Essential Facility

The traditional essential-facilities doctrine generally concerns infrastructure that competitors cannot reasonably duplicate.

In digital markets, the relevant facility may be:

a technical routing infrastructure combined with network effects, data, interoperability, and scale.

The question becomes whether competitors can realistically reproduce the same:

• global network;
• routing intelligence;
• DNS infrastructure;
• Anycast network;
• edge locations;
• telemetry;
• latency measurements;
• customer base;
• interconnection relationships.

If replication is prohibitively difficult, traffic management may acquire essential-facility characteristics.

However, mere technological importance does not automatically satisfy the legal test. Courts generally require careful examination of necessity, duplication, foreclosure, and justification.

5. Relevant Market Definition

Several markets could potentially arise.

Possible upstream markets

• cloud infrastructure;
• CDN services;
• DNS services;
• global traffic-management services;
• network interconnection;
• edge computing;
• DDoS protection.

Possible downstream markets

• online search;
• e-commerce;
• video streaming;
• digital advertising;
• financial services;
• SaaS;
• online gaming.

Possible cross-market theory

A firm could possess dominance in one infrastructure market and use control over traffic routing to favour its own downstream services.

For example:

Dominant cloud infrastructure → global traffic manager → preferential routing → affiliated application

This creates a possible vertical foreclosure theory.

6. Self-Preferencing Through Routing

One of the most important potential concerns is self-preferencing.

A vertically integrated company might operate:

• cloud infrastructure;
• CDN;
• traffic-management system; and
• consumer-facing platform.

Its algorithm could systematically route traffic toward its own infrastructure.

The practice may not appear discriminatory at the contractual level because all competitors technically remain accessible.

But if the algorithm consistently produces:

better latency + greater capacity + higher availability for affiliated services

the competitive effect may resemble preferential access.

Competition authorities would therefore need to examine the actual routing outcomes rather than merely contractual terms.

7. Discriminatory Routing

A dominant routing provider could theoretically give different customers:

• different routing quality;
• different failover treatment;
• different geographical coverage;
• different caching priority;
• different network paths.

This raises questions under discriminatory-abuse theories.

The relevant comparison is not necessarily simply the price paid.

It may involve quality-adjusted access.

For digital infrastructure, discrimination can therefore take the form of:

“same nominal service, different algorithmic performance.”

8. Foreclosure of Rival Infrastructure

Traffic control can also produce network foreclosure.

Assume:

• Provider A operates the dominant traffic-management platform.
• Provider B operates a competing cloud.
• A owns a competing cloud as well.

If A systematically routes traffic away from B, B may experience:

• lower utilisation;
• higher unit costs;
• weaker economies of scale;
• reduced customer attractiveness;
• loss of performance data;
• reduced investment incentives.

This can create a feedback loop:

Traffic → scale → better infrastructure → better routing → more traffic

Once established, such a loop can make digital infrastructure markets highly difficult to contest.

9. Switching Costs and Traffic Lock-In

Global load balancing can reinforce customer lock-in.

A customer may become dependent on:

• proprietary routing rules;
• DNS configurations;
• traffic policies;
• APIs;
• monitoring systems;
• security configurations;
• cloud-specific health checks.

Switching to a competitor can therefore require substantial technical migration.

The result is a form of infrastructure switching cost.

This is particularly important where the customer technically has the right to switch but practically faces:

• downtime risk;
• migration costs;
• performance degradation;
• configuration duplication;
• loss of historical routing data.

10. Data Advantage

Global traffic-management systems generate enormous quantities of information.

A provider may observe:

• where requests originate;
• demand peaks;
• user behaviour;
• network congestion;
• infrastructure failures;
• competitor performance;
• geographic demand;
• application dependency patterns.

This produces a potential data feedback loop:

More traffic → more data → better routing → better performance → more traffic

Consequently, traffic control can produce market power even when the underlying routing technology itself is technically reproducible.

11. Algorithmic Competition Concerns

Modern GLB systems can autonomously optimise traffic.

An algorithm might determine:

“Route 70% of requests to infrastructure A and 30% to infrastructure B.”

If the algorithm incorporates commercially strategic parameters, questions arise regarding:

• discriminatory optimisation;
• algorithmic self-preferencing;
• exclusionary routing;
• coordinated behaviour;
• tacit alignment;
• dynamic capacity allocation.

The competition-law challenge is particularly difficult because the discriminatory rule may be hidden inside a complex machine-learning or optimisation system.

12. Six Important Case Laws

1. United States v. Terminal Railroad Association of St. Louis

This is a foundational case concerning control over infrastructure that competitors needed to reach customers.

The railroad association controlled terminal facilities that were essential for competing railroads entering St. Louis.

Relevance

The case demonstrates how control over a bottleneck infrastructure can have exclusionary consequences when competitors cannot practically compete without access.

Application to global load balancing

A modern analogy could arise where a traffic-management infrastructure becomes indispensable for reaching users or infrastructure efficiently.

The critical issue would be whether the routing facility is genuinely indispensable and whether exclusion substantially restricts competition.

13. 2. Aspen Skiing Co. v. Aspen Highlands Skiing Corp.

The US Supreme Court considered a refusal-to-deal situation involving ski resorts that had historically cooperated through a joint ticketing arrangement.

The Court treated the termination of cooperation as potentially exclusionary because the defendant had sacrificed an existing profitable relationship without an apparent legitimate business justification.

Relevance

The case is important for digital infrastructure because a dominant traffic-management provider may previously have routed traffic across multiple infrastructures and subsequently exclude a rival.

The relevant questions could include:

• Was interoperability previously offered?
• Was cooperation commercially beneficial?
• Was it terminated selectively?
• Was there a legitimate technical justification?

14. 3. Verizon Communications Inc. v. Law Offices of Curtis V. Trinko

The US Supreme Court subsequently adopted a considerably more cautious approach toward compulsory access.

The Court emphasised that antitrust law generally does not require firms to share their infrastructure with competitors merely because access would benefit competition.

Relevance

This is particularly important for global load balancing.

A competitor cannot automatically argue:

“The dominant traffic-management provider has infrastructure, therefore I am entitled to access it.”

The legal analysis must distinguish genuine exclusionary conduct from legitimate independent decisions concerning infrastructure design and investment.

15. 4. Bronner v. Mediaprint

The Court of Justice of the European Union considered refusal of access to a newspaper-delivery system.

The Court applied a stringent test concerning indispensability and whether duplication was realistically possible.

Relevance

The case provides a useful framework for evaluating whether a global routing network constitutes an indispensable facility.

Questions would include:

1. Can the rival create its own routing system?
2. Is alternative traffic management realistically available?
3. Is access indispensable rather than merely advantageous?
4. Would refusal eliminate effective competition?

The indispensability requirement prevents competition law from turning every commercially important infrastructure into a mandatory-access facility.

16. 5. Oscar Bronner / Deutsche Post-type Access Principles

The European competition-law jurisprudence concerning essential infrastructure establishes an important distinction between:

economic convenience and true indispensability.

For global traffic management, a competitor's claim that a particular provider offers the lowest latency would not by itself establish an essential facility.

The competitor would need to demonstrate that alternative infrastructure cannot realistically provide effective competition.

This protects infrastructure operators' incentives to invest while addressing genuine bottleneck exclusion.

17. 6. Google Shopping

The European Commission's Google Shopping case concerned Google's treatment of competing comparison-shopping services in search results.

Although the infrastructure is different from load balancing, the underlying competition principle is highly relevant: a dominant digital intermediary can potentially distort downstream competition through preferential treatment of its own services.

Relevance to traffic control

A similar theory could arise if a dominant traffic intermediary systematically gives its own infrastructure:

• faster routing;
• superior network paths;
• greater capacity;
• preferred failover;
• lower congestion.

The important question is whether control of the intermediary permits the dominant firm to distort competitive opportunities downstream.

18. 7. Microsoft v Commission

The EU Microsoft case is especially important because it involved interoperability and the relationship between a dominant technological platform and adjacent markets.

Microsoft's control over a dominant operating-system environment was relevant to its ability to affect competition in neighbouring software markets.

Relevance

Global load balancing can similarly operate as a technological layer connecting markets.

Control over:

routing → APIs → cloud → applications

may allow a firm to leverage power from one technological layer into another.

19. 8. MEO v Autoridade da Concorrência

The CJEU clarified that discriminatory pricing or conditions under Article 102 TFEU require attention to whether the conduct places trading partners at a competitive disadvantage.

Relevance to routing

This principle can be extended conceptually to non-price discrimination.

A traffic-management provider could potentially discriminate through:

• latency;
• reliability;
• traffic priority;
• geographical access;
• capacity allocation.

Thus, competition analysis should not focus exclusively on monetary charges.

20. Traffic Control and Article 102 TFEU

Under EU competition law, several theories may become relevant.

Article 102(a)

Potentially excessive or unfair access/routing conditions.

Article 102(b)

Limiting technical development or markets.

Article 102(c)

Applying discriminatory conditions to equivalent transactions.

Article 102(d)

Tying or leveraging between infrastructure services.

The precise classification depends upon the conduct and competitive effects.

21. Refusal to Route

A particularly significant scenario would be:

A dominant traffic-management provider refuses to route traffic to a competing cloud or CDN.

This could raise refusal-to-deal concerns.

But the legal test should consider:

• indispensability;
• availability of alternatives;
• duplication possibilities;
• elimination of competition;
• legitimate technical reasons;
• investment incentives;
• security concerns;
• capacity constraints.

A refusal based on genuine network-security or reliability requirements is fundamentally different from exclusion designed to protect an affiliated service.

22. Traffic Throttling

Another potential concern is traffic throttling.

A dominant provider could theoretically:

• reduce routing capacity;
• introduce artificial latency;
• reduce failover priority;
• deprioritise competing services.

The difficulty is evidentiary.

Authorities would need to determine whether the poorer performance resulted from:

legitimate network optimisation

or

strategic exclusion.

This makes technical evidence extremely important.

23. Interoperability as a Competition Remedy

Competition authorities may consider interoperability remedies where a dominant routing platform creates substantial dependency.

Possible remedies include:

1. standardised APIs;
2. transparent routing criteria;
3. portability of routing configurations;
4. interoperability obligations;
5. non-discrimination rules;
6. multi-cloud compatibility;
7. data portability;
8. independent auditing;
9. technical access obligations.

However, remedies must avoid forcing firms to reveal genuine security-sensitive information or proprietary algorithms unnecessarily.

24. Merger-Control Implications

Global load balancing is also relevant to mergers.

Consider a merger between:

• a major cloud provider;
• a leading CDN;
• a DNS provider; and
• a global traffic-management platform.

The transaction could create vertical control over:

user request → routing → edge → cloud → application

Competition authorities could investigate whether the merged firm could:

• foreclose rival clouds;
• favour its own CDN;
• disadvantage competing applications;
• restrict interoperability;
• raise switching costs;
• exploit traffic data.

The concern is therefore not merely horizontal concentration.

It is vertical infrastructure integration.

25. Network Effects

Traffic-control platforms may exhibit strong network effects.

More customers produce:

more traffic data → better routing → better performance → more customers.

This can create cumulative advantages.

Once the network becomes sufficiently large, rivals may face a structural disadvantage even if they possess technically competitive routing technology.

This is sometimes described as a data-performance feedback loop.

26. Multi-Homing as a Constraint

Competition authorities should nevertheless consider whether customers can use multiple routing providers simultaneously.

If customers can easily employ:

• Provider A for DNS;
• Provider B for CDN;
• Provider C for cloud;

then traffic-control power may be weaker.

Multi-homing can provide an important competitive constraint.

But if technical complexity makes multi-homing costly, the nominal availability of alternatives may overstate actual competitive pressure.

27. Security Justifications

A dominant traffic-management provider may legitimately prefer certain infrastructure because of:

• cybersecurity;
• DDoS protection;
• reliability;
• compliance;
• encryption;
• geographic regulations;
• operational stability.

Competition law should not penalise genuine security optimisation.

The critical issue is whether the justification is:

objective, proportionate, consistently applied, and technically verifiable.

A supposed security policy that applies only to competitors could instead indicate discriminatory foreclosure.

28. Regulatory and Jurisdictional Complexity

Global traffic is inherently cross-border.

One routing decision can affect users in:

• India;
• the EU;
• United States;
• Singapore;
• Japan;
• Australia.

Different jurisdictions may therefore investigate the same routing architecture.

This creates problems involving:

• jurisdiction;
• conflicting remedies;
• data localisation;
• cybersecurity regulation;
• telecom regulation;
• competition law;
• cloud regulation.

Global traffic-control systems therefore require increasing coordination between competition authorities and technology regulators.

29. Evidence Required in a Competition Investigation

A sophisticated investigation should examine:

Technical evidence

• routing architecture;
• DNS configurations;
• Anycast topology;
• latency measurements;
• traffic allocation;
• failover rules.

Commercial evidence

• contracts;
• pricing;
• customer migration;
• capacity commitments;
• internal strategy documents.

Algorithmic evidence

• routing objectives;
• optimisation constraints;
• ranking variables;
• exception rules;
• model-training data.

Economic evidence

• diversion ratios;
• switching costs;
• market shares;
• foreclosure effects;
• customer dependency;
• counterfactual routing.

30. Appropriate Competition-Law Framework

A useful analytical sequence is:

Step 1 — Identify the routing layer

What exactly does the company control?

↓

Step 2 — Define the relevant market

Is it cloud, CDN, DNS, traffic management, or a broader market?

↓

Step 3 — Establish market power

Consider market share, network effects, data, switching costs and entry barriers.

↓

Step 4 — Identify the conduct

Self-preferencing, throttling, refusal, tying, discrimination, exclusive dealing, etc.

↓

Step 5 — Determine foreclosure

Are competing providers materially disadvantaged?

↓

Step 6 — Assess indispensability

Can rivals realistically duplicate the routing infrastructure?

↓

Step 7 — Examine legitimate justifications

Security, reliability, capacity, efficiency and technical constraints.

↓

Step 8 — Evaluate competitive effects

Price, quality, innovation, entry, interoperability and consumer choice.

↓

Step 9 — Consider remedies

Non-discrimination, interoperability, portability, transparency or structural remedies where justified.

31. Key Competition-Law Risks

32. Distinguishing Legitimate Optimisation From Anticompetitive Conduct

This distinction is central.

Legitimate optimisation

“Route traffic to the server offering the lowest latency and sufficient capacity.”

Potentially problematic optimisation

“Route traffic away from a rival whenever the rival competes with our downstream service, despite equivalent performance.”

The algorithm itself is not the competition-law violation.

The concern is the competitive objective and effect embedded in its operation.

Conclusion

Global load balancing is evolving from a technical network-management function into a potentially important competitive control point.

The traditional understanding of infrastructure power focused on ownership of physical assets—railways, ports, electricity grids, telecommunications networks and pipelines. In digital markets, equivalent power can increasingly arise from control over the algorithmic allocation of traffic.

The most important legal insight is therefore:

Control over digital traffic can become economically equivalent to control over access to digital markets where users, applications, data and infrastructure depend upon the routing intermediary.

The cases of Terminal Railroad, Aspen Skiing, Trinko, Bronner, Google Shopping, Microsoft and MEO collectively demonstrate the boundaries of this principle. They show, respectively, the importance of bottleneck infrastructure, selective withdrawal of cooperation, limits on compulsory access, indispensability, digital intermediation and self-preferencing, interoperability, and discriminatory treatment.

Nevertheless, global load balancing should not automatically be treated as an essential facility. Competition law must distinguish genuine bottleneck power from ordinary technical superiority. The strongest cases will generally involve a combination of dominance + indispensability or substantial dependency + discriminatory or exclusionary routing + demonstrable competitive harm + weak legitimate justification.

In the emerging digital economy, therefore, competition authorities may increasingly need to treat traffic-routing algorithms, DNS infrastructure, CDN orchestration, edge networks and global traffic managers as economically significant gateways to digital markets, rather than merely as background technical systems.