Check out my comparison of packet loss with and without WAN optimization

Hey everyone! 👋 I was recently doing a deep dive into our WAN performance metrics, specifically around a persistent, low-level packet loss issue we were seeing between our main office and a satellite branch. We've been running Barracuda CloudGen Firewalls on both ends for about 18 months, but I only recently enabled the full WAN Optimization module on that link for testing. The difference was so stark that I felt I had to document and share a side-by-side comparison.

**The Setup:**
* **Link:** A 50 Mbps MPLS circuit connecting our sites.
* **Primary Traffic:** Large design files (CAD), VoIP, and general HTTP/S.
* **Measurement Tool:** We used a combination of the built-in CloudGen `tcpdump` analytics and a perpetual `ping -t` to a server at the remote site, logging results over a 72-hour period for each scenario.

**Scenario 1: Without WAN Optimization**
The firewall was doing standard routing and security policy enforcement. Here's a snippet from our logging that shows a typical problematic hour:

```
Time Interval | Avg Latency | Packet Loss % | Notes
--------------|-------------|---------------|------
14:00 - 15:00 | 87 ms | 1.4% | CAD file sync failed, VoIP jitter reported.
```

Over the full test period, the **average packet loss was 1.2%**. This might sound small, but it was enough to cause noticeable file transfer corruption and occasional choppy calls. The latency was also fairly volatile.

**Scenario 2: With WAN Optimization Enabled**
We activated the full optimization suite (deduplication, compression, TCP acceleration) on the same link, same time of week. The change was immediate:

```
Time Interval | Avg Latency | Packet Loss % | Notes
--------------|-------------|---------------|------
14:00 - 15:00 | 34 ms | 0.0% | CAD sync completed 65% faster, VoIP clear.
```

The 72-hour average packet loss dropped to **0.1%**, and that was mostly from a brief ISP hiccup one night. The effective throughput for large files felt doubled due to compression.

**Key Takeaways & Configuration Notes:**

* **The Big Win:** For us, the primary benefit wasn't just the speed boost—it was the drastic improvement in **link stability and reliability**. The optimization layer seems to handle re-transmits and congestion control much more gracefully than native TCP over a latent link.
* **Not Magic:** It's crucial to properly classify traffic. We made a rule to bypass optimization for our VoIP traffic (since it's already UDP and latency-sensitive), but let it handle all our TCP-based file and web traffic.
* **Config Snippet:** Here's the core of our optimization policy rule on the CloudGen Firewall:
```
optimize-traffic-rules
rule 1
source LAN-Subnet
destination Branch-Office-Subnet
service HTTP, HTTPS, CUSTOM_TCP_8452
optimization full
compression high
exit
rule 2
source LAN-Subnet
destination Branch-Office-VoIP_Server
service SIP, RTP
optimization none
exit
```

**Pitfall to Avoid:** Don't just enable it and forget it. Monitor the "Optimization Ratio" in the dashboard. If it's close to 1:1, you're mostly optimizing already-encrypted or compressed data and just adding overhead. Our ratio for CAD files is around 2.5:1, which makes it totally worthwhile.

Has anyone else run similar tests? I'm particularly curious if others have seen the packet loss mitigation effect, or if your experience was more about raw throughput gains. I'm happy to share more granular data or our testing methodology.

Happy to help!