{"id":19866,"date":"2026-05-13T12:38:39","date_gmt":"2026-05-13T10:38:39","guid":{"rendered":"https:\/\/www.savecall.de\/ki-network-determinism-variance\/"},"modified":"2026-07-30T10:00:29","modified_gmt":"2026-07-30T08:00:29","slug":"ki-network-determinism-variance","status":"publish","type":"post","link":"https:\/\/www.savecall.de\/en\/ki-network-determinism-variance\/","title":{"rendered":"ki network determinism variance"},"content":{"rendered":"\n

\n

\n

Why AI is destroying networks that work on paper<\/h1>\n<\/div>\n\n\n\n

Trusted Advisor for IT & Telecommunications Sourcing<\/h3>\n<\/div>\n\n\n\n
\n
\n
Determinism, variance & data sovereignty: what CIOs need to consider when building AI-enabled networks<\/h2>\n<\/div>\n<\/div>\n\n\n\n
\n
\n
\n
\n
Why variance control & verifiable data paths determine productive AI operation<\/h3>\n\n\n\n
In the architecture documentation, many company networks appear AI-ready<\/a>: high availability, SLAs fulfilled, plenty of bandwidth. In productive operation, however, they still deliver incorrect AI results. The cause is not capacity, but predictability.<\/p>\n\n\n\n
<\/div>\n\n\n\n
In a nutshell:<\/strong><\/p>\n\n\n\n
\n
Variance kills AI, not failure. Jitter, tail latency and brownouts below the SLA threshold are enough to overturn inference and real-time decisions.<\/li>\n\n\n\n
Determinism beats bandwidth. AI workloads need predictable paths and constant latency, not peak throughput.<\/li>\n\n\n\n
Data sovereignty is a question of data flow. If you can’t prove paths in real time, you only have paper.<\/li>\n\n\n\n
Network and security belong in one control level. The old separation no longer supports AI environments.<\/li>\n<\/ul>\n\n\n\n
The solution:<\/strong><\/p>\n\n\n\n
AI-enabled networks<\/a> are built for the worst five minutes of the day, not the monthly average. They connect multiple carriers physically separated, integrate security directly into the data path and make every routing traceable.<\/p>\n\n\n\n
What this article is about: The four specific characteristics of an AI-ready network<\/a>, why classic SLA reports systematically conceal productive AI failures and what architectural decisions IT managers need to make now.<\/p>\n\n\n\n
\n
Book a Consultation<\/a><\/div>\n<\/div>\n<\/div>\n<\/div>\n\n\n\n
\n
\n
Table of contents<\/h3>\n\n\n\n
\n
The end of the network-as-background infrastructure era<\/a><\/li>\n\n\n\n
Speed is the wrong parameter<\/a><\/li>\n\n\n\n
The four network properties that AI really needs<\/a><\/li>\n\n\n\n
Data sovereignty is a data flow problem<\/a><\/li>\n\n\n\n
Conclusion<\/a><\/li>\n\n\n\n
FAQ<\/a><\/li>\n<\/ul>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n\n\n\n\n\n\n\n
\n
\n
The end of the network-as-background infrastructure era<\/h2>\n<\/div>\n<\/div>\n\n\n\n
\n
\n
\n
\n
For decades, the corporate network was regarded as a silent basic service: Move packets, stay out of the way, report availability. Applications failed while the network was technically ‘up’, which was accepted because the damage remained manageable.<\/p>\n\n\n\n
AI fundamentally changes this equation. AI workloads know no idle state. And performance degradation does not lead to a controlled error message, it leads to incorrect outputs, delayed decisions and cascading consequential errors across connected systems. The network is no longer an infrastructure that runs in the background. It is an active determinant of business results.<\/p>\n\n\n\n
At the same time, digital data sovereignty is moving from the compliance agenda to the network architecture. Dependence on large external cloud ecosystems is increasingly becoming a legal and strategic risk variable. Those who cannot control and prove data paths do not have true sovereignty, they only have documentation.<\/p>\n\n\n\n
\n
Book a Consultation<\/a><\/div>\n<\/div>\n<\/div>\n<\/div>\n\n\n\n
\n
$\"Solar$ <\/figure>\n<\/div>\n<\/div>\n<\/div>\n\n\n\n
\n
\n
\n
Speed is the wrong parameter<\/h2>\n<\/div>\n<\/div>\n\n\n\n
\n
\n
\n
$\"Evening$ <\/figure>\n<\/div>\n\n\n\n
\n
\n
Fiber optics for companies: Speed, stability and security<\/h4>\n\n\n\n
The common question ‘Do we have enough bandwidth for AI?’ is too narrow. Bandwidth is necessary, but not sufficient. AI does not fail because data flows too slowly on average. AI fails because data behaves unpredictably.<\/p>\n\n\n\n
The key characteristics of an AI-enabled network are:<\/strong><\/p>\n\n\n\n
\n
\n
1<\/h2>\n<\/div>\n\n\n\n
\n
Determinism:<\/strong><\/p>\n\n\n\n
The network behaves as it should. Not most of the time, but always.<\/p>\n<\/div>\n<\/div>\n\n\n\n
\n
\n
2<\/h2>\n<\/div>\n\n\n\n
\n
Predictable paths:<\/strong><\/p>\n\n\n\n
Routing follows defined rules, not situational optimizations that generate side effects.<\/p>\n<\/div>\n<\/div>\n\n\n\n
\n
\n
3<\/h2>\n<\/div>\n\n\n\n
\n
Consistent latency:<\/strong><\/p>\n\n\n\n
It is not the mean value that counts, but the stability of the value over time.<\/p>\n<\/div>\n<\/div>\n\n\n\n
\n
\n
4<\/h2>\n<\/div>\n\n\n\n
\n
Controlled behavior under load:<\/strong><\/p>\n\n\n\n
The network must not tip over into an uncontrolled state under peak load.<\/p>\n<\/div>\n<\/div>\n\n\n\n
Network design for peak capacity without variance control creates systems that appear stable on paper and fail in production. AI readiness means: design for variance, not for average.<\/strong><\/p>\n\n\n\n
\n
Book a Consultation<\/a><\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n\n\n\n
\n
\n
The four network properties that AI really needs<\/h2>\n<\/div>\n<\/div>\n\n\n\n
\n
\n
Variance, not failure, is the AI killer<\/strong><\/strong><\/h4>\n\n\n\n
Averages lie. Reporting network performance in monthly averages hides the very thing that destabilizes AI systems: the worst five minutes of the day.<\/p>\n\n\n\n
AI workloads are highly sensitive to jitter, packet loss and tail latency. Short brownouts, network degradations below the failure threshold, interrupt inference pipelines and real-time decision-making processes without appearing in classic SLA reports. The result: AI fails silently without the fault being attributed to the network.<\/p>\n\n\n\n
Traditional SLAs smooth performance into key figures that do not reflect the reality of production. Those who design for averages build fragile systems. Those who design for variance build stable ones.<\/p>\n\n\n\n
How procurement logic creates operational fragility<\/strong><\/strong><\/h4>\n\n\n\n
Many corporate networks are optimized for procurement simplicity, not operational reality. A single network provider, a standardized contract, clean SLAs: this makes networks easier to purchase and easier to manage. But performance does not behave uniformly across regions.<\/p>\n\n\n\n
The last mile dominates the quality experienced, and geography plays a decisive role in this. What looks resilient on the architecture diagram becomes fragile under real conditions. Simplicity in purchasing creates fragility in operation.<\/p>\n\n\n\n
When companies scale across countries, clouds and access technologies, performance variance accumulates. Multi-network resilience, rather than dependence on a single last-mile provider, is therefore not a luxury option, but a structural requirement for productive AI deployment.<\/p>\n\n\n\n
“Always On” must be built, not promised<\/strong><\/strong><\/h4>\n\n\n\n
AI workloads penalize even short degradations. Measuring availability after the event is too late, the damage has already been done. Availability must therefore be anchored as an active design decision, not as an SLA promise in the contract.<\/p>\n\n\n\n
In concrete terms, this means:<\/p>\n\n\n\n
\n
Path diversity: Several physically separate access paths as standard, not as an exception.
<\/li>\n\n\n\n
Proactive monitoring: Degradation must be recognized before it affects results.
<\/li>\n\n\n\n
Real-time intervention: Automated response to deviations, not manual escalation processes.<\/li>\n<\/ul>\n\n\n\n
Network and security merge into one control level<\/strong><\/strong><\/h4>\n\n\n\n
AI systems know no rest periods. And neither do attackers. Periodic security checks and subsequent analyses are inadequate for real-time systems. Security must be anchored directly in the data path, continuously, not on an ad hoc basis.<\/p>\n\n\n\n
This means the end of the traditional separation between network and security architecture. In an AI environment, every connection is a real-time security decision. Performance, protection and policy enforcement must operate as one system, not as separate disciplines.<\/p>\n\n\n\n
SASE (Secure Access Service Edge) is the architectural framework that implements this convergence: Network and security under a common control layer, cloud-native, with identity and workload as the control principle instead of geography.<\/p>\n\n\n\n
\n
Book a consultation<\/a><\/div>\n<\/div>\n<\/div>\n<\/div>\n\n\n\n
\n
\n
\n
Data sovereignty is a data flow problem,
not a storage problem<\/h2>\n<\/div>\n<\/div>\n\n\n\n
\n
\n
\n
\n
Checklist: Seven checkpoints for AI-ready networks<\/h4>\n\n\n\n
\n
\n
1<\/h2>\n<\/div>\n\n\n\n
\n
AI data flows mapped end-to-end<\/strong><\/p>\n\n\n\n
Not only for training, but also for inference and retrieval<\/p>\n<\/div>\n<\/div>\n\n\n\n
\n
\n
2<\/h2>\n<\/div>\n\n\n\n
\n
“Always On”<\/strong><\/p>\n\n\n\n
Anchored as a design specification, not formulated as an SLA target<\/p>\n<\/div>\n<\/div>\n\n\n\n
\n
\n
3<\/h2>\n<\/div>\n\n\n\n
\n
Designed for variance<\/strong><\/p>\n\n\n\n
Users experience the worst five minutes, not the average.<\/p>\n<\/div>\n<\/div>\n\n\n\n
\n
\n
4<\/h2>\n<\/div>\n\n\n\n
\n
Data sovereignty<\/strong><\/p>\n\n\n\n
Enforceable via policy and routing, not just documented.<\/p>\n<\/div>\n<\/div>\n\n\n\n
\n
\n
5<\/h2>\n<\/div>\n\n\n\n
\n
Network and security<\/strong><\/p>\n\n\n\n
Converged into a common control level.<\/p>\n<\/div>\n<\/div>\n\n\n\n
\n
\n
6<\/h2>\n<\/div>\n\n\n\n
\n
End-to-end instrumentation available:<\/strong><\/p>\n\n\n\n
Behavior is visible and verifiable.<\/p>\n<\/div>\n<\/div>\n\n\n\n
\n
\n
7<\/h2>\n<\/div>\n\n\n\n
\n
Multi-network resilience built:<\/strong><\/p>\n\n\n\n
No dependence on a single last-mile provider.<\/p>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n\n\n\n
\n
\n
Sovereignty is a question of data flow<\/h4>\n\n\n\n
Data sovereignty is often treated as a question of storage location. This falls short. Sovereignty is not defined by where data is stored, but by how data flows.<\/p>\n\n\n\n
Modern network architectures for AI workloads must make data flows observable and verifiable. This means that companies must be able to document the paths through which data has flowed, under which policies and with which access rights, not as a retrospective reconstruction, but in real time.<\/p>\n\n\n\n
Anyone who cannot observe the path has no sovereignty. They have paper documentation. This is particularly critical in regulated environments such as EMEA, where regulatory expectations go far beyond data retention and require operational traceability through EU AI Act, NIS2 and industry-specific requirements.<\/p>\n\n\n\n
<\/div>\n\n\n\n
\n
Book a Consultation<\/a><\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n\n\n\n
\n
\n
Conclusion: CIOs don’t have a bandwidth problem, they have a predictability problem<\/h2>\n<\/div>\n\n\n\n
\n
\n
The real challenge for AI in production is not a lack of capacity. It is a lack of predictability. Networks can no longer be passive infrastructure, they must function as an intelligent, controllable layer that supports continuous change without friction. If you are serious about scaling AI workloads, you need a network design that focuses on determinism, variance control and data sovereignty, and an operational approach that keeps pace with the dynamics of AI.<\/p>\n\n\n\n
How SAVECALL supports you<\/h4>\n\n\n\n
As a carrier-neutral sourcing and consulting partner with over 80 carrier partnerships worldwide, SAVECALL supports IT managers in setting up AI-capable network architectures, from requirements analysis to provider and price comparison to ongoing operational management.<\/p>\n\n\n\n
We assess your existing WAN for determinism and variance resilience, identify last-mile dependencies and work with you to design an underlay that supports AI workloads, with SASE integration, multi-carrier resilience and verifiable data paths.<\/p>\n\n\n\n
\n
Get free advice now<\/a><\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n\n\n\n
\n
\n
\n
\n
$\"Frank$ <\/figure>\n<\/div>\n\n\n\n
\n
Written by<\/p>\n\n\n\n
Frank Frommknecht<\/h3>\n\n\n\n
Key Account Consultant, SAVECALL<\/strong><\/p>\n\n\n\n
Has supported companies for over 20 years in selecting and optimizing their connectivity solutions. His focus: making complex telecommunications understandable from the customer’s perspective and finding the right strategic solution.<\/p>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n\n\n\n
\n
\n
$\"SAVECALL$ <\/figure>\n\n\n\n
Kostenloser<\/span> Expertentermin<\/a><\/h3>\n\n\n\n
\n
Buchen<\/a><\/div>\n<\/div>\n<\/div>\n<\/div>\n\n\n\n
\n
\n
Sources<\/h3>\n\n\n\n
\n
IDC \/ Network World: AI in Networking Special Report 2026<\/a><\/li>\n\n\n\n
Cisco AI Readiness Index<\/a><\/li>\n<\/ul>\n<\/div>\n<\/div>\n\n\n\n
\n
\n
Frequently Asked Questions About AI-Enabled Networks<\/h2>\n<\/div>\n<\/div>\n\n\n\n
Frequently Asked Questions About AI-Enabled Networks<\/h3>\n\n\n\n
Why is variance more dangerous for AI than a complete outage?<\/strong>
An outage is detectable and gets escalated. Variance, meaning jitter, tail latency and short brownouts below the SLA threshold, stays invisible in classic monthly averages, but still overturns inference results and real-time decisions. AI fails silently as a result, without the fault ever being attributed to the network.<\/p> <\/div>
Why isn’t sufficient bandwidth enough for productive AI operation?<\/strong>
Bandwidth is necessary, but not sufficient. AI workloads need predictable paths and constant latency rather than pure peak throughput. A network with plenty of bandwidth but fluctuating performance still delivers unstable AI results, because determinism decides success, not capacity.<\/p> <\/div>
What does data sovereignty concretely mean in a network context?<\/strong>
Data sovereignty is not a question of storage location, but of data flow. Companies must be able to prove in real time which paths data flows through, under which policies and with which access rights. If you cannot observe the path, you only have paper documentation instead of true sovereignty.<\/p> <\/div>
Why does simple procurement often create fragile networks?<\/strong>
A single network provider with a standardized contract and clean SLAs makes purchasing easier, but performance does not behave uniformly across regions. The last mile dominates the experienced quality. Multi-network resilience, instead of dependence on a single last-mile provider, is therefore a structural requirement for productive AI deployment.<\/p> <\/div>
Why do network and security merge into one control level?<\/strong>
AI systems know no rest periods, and neither do attackers. Periodic security checks are not enough for real-time systems. Every connection is a real-time security decision, which is why performance, protection and policy enforcement must operate as one system. SASE provides the architectural framework for this.<\/p> <\/div>
How does SAVECALL support building AI-enabled networks?<\/strong>
As a carrier-neutral sourcing and consulting partner, SAVECALL assesses your existing WAN for determinism and variance resilience, identifies last-mile dependencies and works with you to design an underlay that supports AI workloads, with SASE integration, multi-carrier resilience and verifiable data paths.<\/p> <\/div> <\/div>\n\n\n\n
\n
\n
Articles that may also interest you<\/h2>\n<\/div>\n\n\n\n
\r\n
\r\n
\r\n \r\n\r\n
\r\n
\r\n $\"SAVECALL$ <\/div>\r\n
Security<\/p>\r\n <\/div>\r\n\r\n
\r\n
managed-security-service-provider-germany<\/h3>\r\n
\r\n
Learn moreLearn more<\/h3>\r\n <\/div>\r\n <\/div>\r\n\r\n <\/a>\r\n <\/div>\r\n
\r\n \r\n\r\n
\r\n
\r\n $\"Satellite$ <\/div>\r\n
WAN<\/p>\r\n <\/div>\r\n\r\n
\r\n
wan-costs-site-connectivity<\/h3>\r\n
\r\n
Learn moreLearn more<\/h3>\r\n <\/div>\r\n <\/div>\r\n\r\n <\/a>\r\n <\/div>\r\n
\r\n \r\n\r\n
\r\n
\r\n $\"Satellite$ <\/div>\r\n
WAN<\/p>\r\n <\/div>\r\n\r\n
\r\n
wan-monitoring-visibility-operation<\/h3>\r\n
\r\n
Learn moreLearn more<\/h3>\r\n <\/div>\r\n <\/div>\r\n\r\n <\/a>\r\n <\/div>\r\n
\r\n \r\n\r\n
\r\n
\r\n $\"Satellite$ <\/div>\r\n
WAN<\/p>\r\n <\/div>\r\n\r\n
\r\n
wan-sla-mttr-what-counts<\/h3>\r\n
\r\n
Learn moreLearn more<\/h3>\r\n <\/div>\r\n <\/div>\r\n\r\n <\/a>\r\n <\/div>\r\n
\r\n \r\n\r\n
\r\n
\r\n $\"Satellite$ <\/div>\r\n
Voice<\/p>\r\n <\/div>\r\n\r\n
\r\n
cloud-phone-system-provider-comparison<\/h3>\r\n
\r\n
Learn moreLearn more<\/h3>\r\n <\/div>\r\n <\/div>\r\n\r\n <\/a>\r\n <\/div>\r\n
\r\n \r\n\r\n
\r\n
\r\n $\"SAVECALL$ <\/div>\r\n
Mobile radio management<\/p>\r\n <\/div>\r\n\r\n
\r\n
Business Mobile Provider Comparison 2026<\/h3>\r\n
\r\n
Learn moreLearn more<\/h3>\r\n <\/div>\r\n <\/div>\r\n\r\n <\/a>\r\n <\/div>\r\n
\r\n \r\n\r\n
\r\n
\r\n $\"SAVECALL$ <\/div>\r\n
Internet<\/p>\r\n <\/div>\r\n\r\n
\r\n
Business Internet Connection Which Provider Fits<\/h3>\r\n
\r\n
Learn moreLearn more<\/h3>\r\n <\/div>\r\n <\/div>\r\n\r\n <\/a>\r\n <\/div>\r\n
\r\n \r\n\r\n
\r\n
\r\n $\"SAVECALL$ <\/div>\r\n
Consulting<\/p>\r\n <\/div>\r\n\r\n
\r\n
why-vendor-sprawl-destroys-agility<\/h3>\r\n
\r\n
Learn moreLearn more<\/h3>\r\n <\/div>\r\n <\/div>\r\n\r\n <\/a>\r\n <\/div>\r\n
\r\n \r\n\r\n
\r\n
\r\n $\"Das$ <\/div>\r\n
WAN<\/p>\r\n <\/div>\r\n\r\n
\r\n
AI-Enabled Networks: Why Bandwidth Alone Isn’t Enough\u2014and What CIOs Really Need to Focus On<\/h3>\r\n
\r\n
Learn moreLearn more<\/h3>\r\n <\/div>\r\n <\/div>\r\n\r\n <\/a>\r\n <\/div>\r\n
\r\n \r\n\r\n
\r\n
\r\n $\"SAVECALL$ <\/div>\r\n
WAN<\/p>\r\n <\/div>\r\n\r\n
\r\n
Global Connectivity from a Single Source<\/h3>\r\n
\r\n
Learn moreLearn more<\/h3>\r\n <\/div>\r\n <\/div>\r\n\r\n <\/a>\r\n <\/div>\r\n
\r\n \r\n\r\n
\r\n
\r\n $\"SAVECALL$ <\/div>\r\n
Security<\/p>\r\n <\/div>\r\n\r\n
\r\n
nis2-telecom-infrastructure<\/h3>\r\n
\r\n
Learn moreLearn more<\/h3>\r\n <\/div>\r\n <\/div>\r\n\r\n <\/a>\r\n <\/div>\r\n
\r\n \r\n\r\n
\r\n
\r\n $\"WAN$ <\/div>\r\n
WAN<\/p>\r\n <\/div>\r\n\r\n
\r\n
managed-sd-wan-provider-germany<\/h3>\r\n
\r\n
Learn moreLearn more<\/h3>\r\n <\/div>\r\n <\/div>\r\n\r\n <\/a>\r\n <\/div>\r\n
\r\n \r\n\r\n
\r\n
\r\n $\"SAVECALL$ <\/div>\r\n
Internet<\/p>\r\n <\/div>\r\n\r\n
\r\n
sase-provider-comparison<\/h3>\r\n
\r\n
Learn moreLearn more<\/h3>\r\n <\/div>\r\n <\/div>\r\n\r\n <\/a>\r\n <\/div>\r\n
\r\n \r\n\r\n
\r\n
\r\n $\"SAVECALL$ <\/div>\r\n
Voice<\/p>\r\n <\/div>\r\n\r\n
\r\n
Why cloud security is important<\/h3>\r\n
\r\n
Learn moreLearn more<\/h3>\r\n <\/div>\r\n <\/div>\r\n\r\n <\/a>\r\n <\/div>\r\n
\r\n \r\n\r\n
\r\n
\r\n $\"Savecall$ <\/div>\r\n
Partner<\/p>\r\n <\/div>\r\n\r\n
\r\n
Second acquisition in twelve months: Savecall takes over landline customers from Bluestring<\/h3>\r\n
\r\n
Learn moreLearn more<\/h3>\r\n <\/div>\r\n <\/div>\r\n\r\n <\/a>\r\n <\/div>\r\n
\r\n \r\n\r\n
\r\n
\r\n $\"SAVECALL$ <\/div>\r\n
Internet<\/p>\r\n <\/div>\r\n\r\n
\r\n
What and where are the challenges for the network infrastructure of most companies?<\/h3>\r\n
\r\n
Learn moreLearn more<\/h3>\r\n <\/div>\r\n <\/div>\r\n\r\n <\/a>\r\n <\/div>\r\n
\r\n \r\n\r\n
\r\n
\r\n $\"SAVECALL$ <\/div>\r\n
Internet<\/p>\r\n <\/div>\r\n\r\n
\r\n
SAVECALL Holding GmbH founded: Clear structure. Strong foundation. Long-term partnership.<\/h3>\r\n
\r\n
Learn moreLearn more<\/h3>\r\n <\/div>\r\n <\/div>\r\n\r\n <\/a>\r\n <\/div>\r\n
\r\n \r\n\r\n
\r\n
\r\n $\"SAVECALL$ <\/div>\r\n
Security<\/p>\r\n <\/div>\r\n\r\n
\r\n
Mastering Shadow AI with SASE: Security and transparency for Gen-AI in the company<\/h3>\r\n
\r\n
Learn moreLearn more<\/h3>\r\n <\/div>\r\n <\/div>\r\n\r\n <\/a>\r\n <\/div>\r\n
\r\n \r\n\r\n
\r\n
\r\n $\"Luftaufnahme$ <\/div>\r\n
Partner<\/p>\r\n <\/div>\r\n\r\n
\r\n
10 advantages of mobile device management<\/h3>\r\n
\r\n
Learn moreLearn more<\/h3>\r\n <\/div>\r\n <\/div>\r\n\r\n <\/a>\r\n <\/div>\r\n
\r\n \r\n\r\n
\r\n
\r\n $\"Modern$ <\/div>\r\n
Internet<\/p>\r\n <\/div>\r\n\r\n
\r\n
Insights into the requirements of a modern workplace<\/h3>\r\n
\r\n
Learn moreLearn more<\/h3>\r\n <\/div>\r\n <\/div>\r\n\r\n <\/a>\r\n <\/div>\r\n <\/div>\r\n\r\n
<\/div>\r\n
<\/div>\r\n
<\/div>\r\n<\/div>\r\n\r\n