Hey folks, I know this is a bit outside our usual CI/CD wheelhouse, but as someone who's obsessed with infrastructure stability and performance metrics, I've been living with a SonicWall TZ500 as my edge device for exactly three years now. I'm curious if others have seen what I'm starting to document: a subtle but perceptible performance degradation over the long haul.
When I first racked it, the throughput was rock-solid for our needs. We're a dev-heavy shop, so our traffic is a mix of SSH, massive artifact transfers (think Docker images and pipeline caches), and a ton of web traffic for dependency downloads. The initial setup was a dream, and for the first 18-24 months, it was a "set it and forget it" piece of kit. Lately, though, I've been noticing the need for more frequent reboots—maybe once every 6-8 weeks—to clear what feels like a "sluggish" state, especially during concurrent high-load scenarios like:
* Full-team pushing to repos at 5 PM.
* Parallel security scans kicking off in multiple pipelines.
* Large, multi-gigabyte cloud syncs.
I've been logging some numbers. Here's a snippet from a simple `iperf3` test I run weekly between our build server and an external host (through the firewall, of course). The drop isn't catastrophic, but it's there:
```bash
# Month 6 (Initial Baseline)
[ ID] Interval Transfer Bitrate
[ 5] 0.00-10.00 sec 1.15 GBytes 985 Mbits/sec
# Month 30 (Recent)
[ ID] Interval Transfer Bitrate
[ 5] 0.00-10.00 sec 987 MBytes 828 Mbits/sec
```
My configuration has gotten more complex over time, sure. I've added:
* Deep Packet Inspection for specific subnets.
* Geo-IP filtering.
* More granular App Control policies for SaaS tools.
* Constant firmware updates (now on 6.5.4.9-94n).
So, my big question to the community: **Is this just the expected outcome of layering on features and living with a device for years, or have others experienced a genuine "hardware fatigue" or memory leak-type scenario with the TZ series?** I'm trying to decide if this is a call to support, a hardware refresh signal, or just the natural state of a busy firewall that needs a config audit and trim. From a CI/CD perspective, unreliable or throttled network is a silent pipeline killer!
pipeline all the things
You're chasing a hardware problem with iperf tests. It's not the throughput. Check memory fragmentation and the session table.
Our TZ600 started dropping random UDP packets after 2.5 years, specifically during heavy artifact syncs. The logs showed nothing. A hard factory reset and config rebuild solved it for another year. Their stateful inspection engine seems to leak over time.
What's your session count look like during those 5 PM pushes? If you're hitting 80% of the limit, that's your culprit. Reboots clear the table.
slow pipelines make me cranky
> "I've been logging some numbers. Here's a snippet from a simple `iperf3` test..."
I'd be cautious about calling iperf3 "simple" when you're testing through a stateful firewall. The default iperf3 TCP test uses a single stream and a very small window, which tells you almost nothing about the device's ability to handle concurrent sessions or deep-packet inspection under load. You're measuring the wire, not the appliance's real-world state.
What matters far more is the session table occupancy and the memory fragmentation that user344 mentioned. I'd be surprised if you're not seeing a steady climb in "concurrent connections" in the SonicWall dashboard after weeks of uptime, especially with Docker pulls and pipeline scans generating thousands of short-lived TCP connections. A reboot resets the table, which is why the "sluggishness" clears temporarily.
If you want a reproducible benchmark, set up a test with multiple parallel iperf3 streams (try `-P 16` or 32) and run it right after a cold boot, then again after 45 days of uptime. Also capture the `DPI SSL` CPU usage and the connection table utilization at the same time. Without that, you're describing a feeling, not a defect.
Have you tried setting a session timeout lower than the default 5 minutes? That alone can keep the table from bloating during heavy artifact transfers.
Trust but verify.