Region-specific matters in product design and cloud architecting.
- Charlotte Letamendia
- Mar 15, 2024
- 9 min read
Updated: Jul 19
That is a no brainer that low latency is crucial for lots of applications, especially in media or gaming with voice and video based real-time communication.
Let's drive deep together the choice of cloud regions when we are in the design of a product. We are helping product manager to make informed decisions without compromising cost, performance and roadmap.
The article is based on an open dataset and open architecture reference designs.
Our story telling follows the use case of a media company based in Argentina developing an application hosted in the cloud and with end users all over Latin America.
Question 1 - When localization matters ?
Localization matters when
latency matters
data residency matters
cost matters
product catalog matters
1 - when latency matters
When responsiveness for end users is important.
In the delivery of web content such as videos, webpages, and images to serve to a application, a low latency improve the user experience. The faster data are served, the better is the user experience.
Latency is the delay in sending information from one point to the next; it affects response times. It’s reduced when processing at the edge because data produced by sensors and IoT devices no longer needs to be sent to a centralized cloud to be processed.
Servers geographically closer to the users guarantee lower latency and avoid network variation and intermediary layer and speed up application loading for data heavy applications.
Hosting the servers, front and back, in a distant region increases communication traffic between the application and its users as data must travel over long physical distances through the complexity of internet. The communication is also two way, with requests going from the client to the server and responses coming back. For example, when a user visits a website, data from that website's server has to travel across the internet to reach the user's computer. If the user is located far from that server, it will take a long time to load a large file, such as a video or website image.
Below 3 referenced design architecture showing the need of low latency
Media use case
All common media and web application use static content stored at edge servers geographically closer to the users such as CDN. These files
can include static content, such as HTML, CSS, JavaScript, image, and video files.
.
Database use case
To serve data from dynamic sources such as databases or real-time computations, edge servers can be used to provide caching or deploy dedicated compute instances in local regions. As shown here in the referenced architecture design.
Inference use case
The application use an AI algorithm that retrain and learn from the actions, and provide contextual action, images, speech, characters. The application call an application with IA inference where the response is expected with les than ms. Moreover the AI algorithm retrains and learns from the actions
Example: photorealistic neural rendering, automating storytelling with large language model, automatically programing portion of the games, realistic NPCs (Non-Playable Characters) with LLM.
2- When data residency matters
Understanding and complying with these laws or compliance by country, nationality, sector of activity and type of data involved is essential for organizations as non-compliance can lead to significant legal and financial consequences.
Identifying the rules your service is subject too
Project the design impact and technical requirements for the design
The rules
Different countries and regions have distinct legal and regulatory influence.
When developing a service, first map the regulation you are subject to:
The country where the data is hosted (influence1)
ex: Cloud act, NCC
Nationality of the end user (influence2)
ex: RGPD
The country of the service provider (influence3)
ex: Cloud act
The sector of activity (influence4)
ex: Health and financial industry
The type of data involved (influence5)
ex: Health and financial industry
The requirements for the cloud provider
Compliance and data protection laws require cloud designs to prioritize privacy, security, transparency, and control mechanisms at every architectural layer, fundamentally influencing technical decisions and operational processes, such as :
Feature list required from a cloud provider to comply with data protection laws
✅ Security controls
encryption (at rest and in transit)
access management (such as role-based access control)
✅ Data Localization and Transfer Restrictions
storage restriction (processed and stored)
data tracking
transparency about data flows
✅ Auditing and Monitoring
compliance frameworks
audit and monitoring reports
maintained detailed logs
3- When cost of bandwidth matters
Bandwidth is the rate at which data is transferred over the internet. When data is sent to the cloud, it travels through a wide area network, which can be costly due to its global coverage and high bandwidth needs. When processing data at the edge, local area networks can be utilized, resulting in higher bandwidth at lower costs.
A POP in Buenos Aires ensures your data gets onto a high-speed, reliable network as close as possible to your source, reducing latency and increasing transfer reliability. But it has a cost.
A dedicated private network connection from your premises directly to the cloud provider is a dedicated connection that provides consistent, low-latency, high-bandwidth, and stable network performance. Ideal for enterprises with large, ongoing data transfer needs requiring stability and security. Requires physical setup, provisioning, and possibly contracts with network providers. Private connection, reducing exposure to internet threats. For steady, large-scale, sensitive, or latency-critical transfers. It has a cost too.
Alternative is to use a public internet transfer service that routes data through from edge locations to the cloud region.
Back up of 10 TB every day from a private on premise cloud in Buenos Aires (Argentina) to AWS East 1A in US
Cost Component | AWS Direct Connect (1 Gbps) | AWS S3 Transfer Acceleration (Internet) |
Port Fee (monthly) | ~$216 (0.30/hr × 24 × 30) See price here | N/A |
Data Transfer Fee (monthly) | ~$6,144 (300,000 GB × $0.02/GB) See price here | ~$24,000 (300,000 GB × $0.08/GB) See price here |
Acceleration Fee (monthly) | N/A | Included above (S3 TA: $0.08/GB from S. America) |
Total (monthly) | ~$6,360 | ~$24,000 |
By staying close, this cost decreases. Within the same region, the Data Transfer Fee drops to zero and the Egress Fee is zero, so the data migration cost is almost negligible (only the port rental fee of a few hundred dollars per month).
4- When cost of service matters, storage example
The cost of storage is regional dependent.
The cost depends on the location of a datacenter heavily influences storage costs:
✅ Electricity costs Datacenters consume a huge amount of power, both for servers and for cooling. Energy prices vary significantly by country or even region. For example, hosting in a place with cheap hydro or nuclear power (e.g., parts of Scandinavia) can be much cheaper than places relying on expensive fossil fuels.
✅ Real estate and construction costs Urban centers or regions with high land and labor costs (like Tokyo, New York, or London) make building and operating datacenters more expensive compared to rural or less developed areas.
✅ Cooling needs & climate Colder climates allow datacenters to use “free cooling” —drawing in outside air — reducing cooling costs. Hot, humid places require expensive air conditioning, increasing operational expenses.
✅ Network connectivity & bandwidth The cost of high-speed internet backbones and peering arrangements varies. In some regions, network transit fees are much higher, impacting storage costs (since moving data in/out is part of the storage cost model).
✅ Local taxes & regulations Some countries/states incentivize datacenters with tax breaks, while others impose heavy taxes, carbon levies, or strict data residency requirements, which can increase costs.
✅ Security & political risk Regions with instability or higher risk may require additional investment in physical and cybersecurity, insurance, or redundancy, driving up the price.
✅ Demand & competition If there’s high demand but few datacenter providers (low competition), prices stay high. Conversely, regions with many datacenters (like Northern Virginia or Frankfurt) see competitive pricing.
So, two datacenters with the same hardware could offer very different storage pricing purely based on where they’re located and the costs tied to that location.
As I need for my game application to store huge amount of videos and reduce the latency I am looking for a region nearby Buenos Aires.
Argentina : expensive
Brazil : expensive
Chili : cheap
Price for Standard tier of Object Storage
Pays | AWS ($/GB/mois) | GCP ($/GB/mois) | Azure ($/GB/mois) | Huawei Cloud ($/GB/mois) |
Brésil (São Paulo) | 0.046 | 0.046 | 0.0476 | 0.045 |
Chili (Santiago) | 0.0475 | 0.047 | 0.048 | Non disponible |
Argentine (Buenos Aires) | Non disponible | 0.047 | Non disponible | Non disponible |
Colombie (Bogota) | Non disponible | 0.047 | 0.048 | Non disponible |
Price in latam
Price variation: between +10% and +40% (Huawei (-15%) is an exception)
Analysis
In Latin America, cloud services are available in Brazil and, more recently, in Chile. The main providers include the usual players: GCP, Oracle, Huawei, AWS, and Azure..
Brazil is significantly more expensive compared to other regions due to local taxes, and it primarily serves the local market, making it less suitable for foreign markets.
Except for Oracle, no provider offers the two regions in the same country
Compromise?
There’s no compromise on price — Brazil is excluded from the list of eligible options.
Using two regions with one provider offers simplicity: a single configuration, a single contract. On the other hand, choosing two regions across two providers aligns with best practices for resilience and allows better leverage for price negotiation, even if it requires additional effort to support multi-cloud deployment.
Simplicity or flexibility?I choose flexibility.
Therefore, my recommendation is to designate AWS (AR-US) as the primary deployment zone and GCP (Chile) as the secondary, redundant zone.
Comparaison of prices between regions by provider
based on the price of one service : Standard Object Storage price (Go/month) in LATAM
compared to USA North (N. Virginia) : 0,023 USD/Go/month
4- When catalog matters
Developing games i have in between others 3 specific requirements :
Zoom in the components in AWS local region in Buenos Aires
Today only AWS local zone is offering a GPU option in Buenos Aires.
The C5 group of instances is suitable for my applications as it is design for online gaming, machine learning, media transcoding: compute-intensive tasks
The G4dn instances are the best option for small-scale machine learning (ML) training and GPU-based ML inference
The R5 Instances for memory-intensize applications, such as mid-size databases and real-time big data analytics.
Compromize ? Lucky me, some GPU are available in AWS local zone in Argentina
>>> Take away
As a product manager working with architect,
I am informed and take unbiased decisions. I found the best compromise to improve performance and reduce costs, based on data and technical tests.
It is not easy to have a good and unique solution. As a result we did small compromises:
Moderate compromise: A slight increase in latency (a few nanoseconds) by choosing hosting in Santiago (GCP).
No compromise on cost: Hosting the application in Buenos Aires to optimize expenses.
No compromise on Infrastructure as Code: Ensuring agility and speed in the roadmap.
The Argentina landscape is specific due to local constrains such as inflation and importation and usual patterns of implementation and design are not plugeable for argentina deployment. Compromize have to be done and in this contexte proabably hybrid architecture are necessary to mix advantage of the cloud and edge servor to bring latency.
Lasnubesalternativas helps you make informed and unbiased decisions to improve performance and reduce costs, based on data and technical tests.
Question 2 - How to do with a good trade off
As a product manager, you often have to make trade-offs between performance, cost, and feasibility.
In Argentina a large numbers of providers is offering Virtual Private Servers. This servers are available in a pay-as-you-go mode and can be configured and managed by users according to their specific needs.
In the opposite only a few providers are offering a complete suite of product enabling automation, management, and deployment of infrastructure through code instead of manual configurations. This suite is proposed by main cloud provider, on the contrary of VPS and DS providers.
So that the infra is repeatable, scalable, and consistency. Today, at the edge of AI coding and agent for deployment and decision, cloud services appears as a must to deploy our services and gain in team frugality, and velocity on our roadmap.
Table 1 – Breakdown of Service Providers by Service Type in Buenos Aires
Service | Number of Providers | Providers | |
0 | Cloud | 1 | Huawei |
1 | Cloud LZ | 1 | AWS |
2 | DC | 1 | |
3 | Network | 1 | Sencinet |
4 | VPS | 14 | Cirion, SyT Datacenter, IPXON, SkyOnline, Giga... |
5 | Web | 1 |
Analysis
x 14 Providers are offering VPS.
x 2 Cloud providers
Building over VPS is a more complex alternative than over a complete cloud product line.
Nearby regions such as São Paulo (Brazil) and Santiago (Chile) also offer viable alternatives — whether for primary deployment or for enhancing regional resiliency.
Table 2 – Breakdown of Cloud Service Providers by in Latam
City | Number | Providers | |
0 | Bogota | 1 | [Oracle] |
1 | Buenos Aires | 2 | [Huawei, AWS LZ] |
2 | Lima | 1 | [Huawei] |
3 | Rio de Janeiro | 1 | [Azure] |
4 | Santiago | 3 | [Oracle, Huawei, GCP ] |
5 | Sao Paulo | 6 | [Oracle, Linode, AWS, Azure, Huawei, GCP] |
6 | Valparaiso | 1 | [Oracle] |
7 | Vinhedo | 1 | [Oracle] |
Data in latam
14 regions
1 local zone
>> Chili and Brazil can be alternative looking for a region nearby Buenos Aires
Comments