> For a complete page index, fetch https://docs.synthflow.ai/llms.txt. For full documentation content, fetch https://docs.synthflow.ai/llms-full.txt.

# Deploy in LATAM Regions

> Real-world examples of regional deployment and backbone routing improvements.

We are able to deploy in LATAM regions to reduce latency and improve call quality for customers in Latin America.

## Why regional deployment helps

Placing media and signaling close to callers reduces **RTT** and **one‑way latency**.\
Shorter paths also reduce **PDV/jitter** and **loss** by avoiding congested or long‑haul links.

## Private backbone routing

When traffic must cross regions, routing over the **Synthflow backbone** (instead of the public internet) can:

* Reduce hop count and variable queuing delays.
* Stabilize jitter by avoiding congested public segments.
* Improve loss characteristics under peak loads.

> Note\
> Actual results depend on last‑mile access and your carrier. Validate with the measurement steps in the case study below.

## Regionalization playbook

* Start with current call distribution and busiest geographies.
* Measure baseline metrics (see performance metrics page).
* Request a **regional PoP** deployment near the callers.
* Re‑route calls/media to the new region or over the Synthflow backbone.
* Re‑measure and compare before/after.

## Case study — LATAM regionalization

**Background**\
A major customer in **LATAM** experienced variable quality when traffic egressed via **US** infrastructure during peak hours.

**Problem**\
Not only higher latency, but also **packet loss** and **packet delay variation (PDV/jitter)** led to choppy/robotic speech and occasional brief “squeal” during bursty delivery.

**Intervention**\
We deployed a **LATAM** point of presence and routed inter‑region traffic over the **Synthflow backbone** instead of the public internet.

**Results**\
Significant improvement in **RTT (end‑to‑end)** and overall call quality.

**Before/After metrics**

| Metric               | Before (US) | After (LATAM) |           Δ | Notes               |
| -------------------- | ----------: | ------------: | ----------: | ------------------- |
| RTT e2e (ms)         |   **191.3** |      **99.5** |    **−48%** | Measured end‑to‑end |
| One‑way latency (ms) |    **95.7** |      **49.8** |    **−48%** | p50 / p95           |
| PDV/jitter (ms)      |  **\[TBD]** |    **\[TBD]** | **\[TBD]%** | p95                 |
| Packet loss (%)      |  **\[TBD]** |    **\[TBD]** | **\[TBD]%** | p95                 |
| MOS                  |  **\[TBD]** |    **\[TBD]** |  **\[TBD]** | E‑Model             |
| Call setup time (s)  |  **\[TBD]** |    **\[TBD]** | **\[TBD]%** | 100→200 total       |

**Measurement methodology**

* **RTT/Path:** `mtr -u -c 200 <media-endpoint>` and `ping -c 200 <media-endpoint>`.
* **RTP stats:** Capture a sample of **≥100 calls** with Wireshark; use **RTP Stream** analysis for jitter/loss/skew.
* **Synthetic traffic:** `sipp` for call timing; optional `iperf3 -u` for UDP jitter/loss characterization when a test server is available.
* **MOS:** Compute using E‑Model from measured one‑way delay, jitter, and loss.

**What to watch for**

* ISP changes and peering shifts that re‑introduce jitter.
* Packet bursts at hour boundaries (dialer starts) causing short‑term PDV spikes.
* SBC/NAT keep‑alive intervals.

**Replication checklist**

* Confirm call distribution to LATAM PoP.
* Enable backbone routing for inter‑region segments.
* Verify codecs (start with G.711).
* Tune jitter buffer on PBX/SBC if configurable.
* Re‑measure and compare p50/p95/p99 against baseline.

### Metric definitions

| **Metric**                    | **Meaning**                                                                                         |
| ----------------------------- | --------------------------------------------------------------------------------------------------- |
| **RTT e2e (Round-Trip Time)** | How long it takes for your voice signal to go to the other person and back — like the “echo delay.” |
| **One-way latency**           | The delay from when you speak to when the other person hears you.                                   |
| **PDV / jitter**              | How much that delay changes over time — if it jumps around, voices can sound choppy.                |
| **Packet loss**               | How many bits of voice data get lost in transit — too much loss makes speech cut out.               |
| **MOS (Mean Opinion Score)**  | A 1–5 quality score for how clear the call sounds overall.                                          |
| **Call setup time**           | How long it takes after you dial before the call starts ringing or connects.                        |

> Real-world examples of regional deployment and backbone routing improvements.