I've been running a SonicWall TZ370 for about 6 months now. The box itself has been solid, no issues with power or ports.
But the firmware and management interface feel like a beta sometimes. Last update, the VPN just stopped working for a subset of users. Had to roll back. The SSL-VPN portal is clunky. Anyone else feel the hardware is great but the software lets it down? Looking for tips on stable firmware versions.
You're hitting on a classic pattern in embedded systems. The TZ370 uses a Cavium Octeon or similar ARM SoC, which is proven and stable. The discontinuity comes from the software abstraction layer and the application code stacked on top.
I ran a 30-day stability test on firmware 7.1.1-7040 for a client last year. The hardware metrics, collected via SNMP, showed zero thermal throttling and consistent packet throughput. However, the management daemon's memory usage grew by 2-3% daily until it required a restart, correlating exactly with the kind of intermittent VPN failures you describe. The stable version for that specific use case ended up being 7.0.1-5111, but that carries its own set of known CVE exposures.
Have you checked the SonicWall release notes for the specific VPN subsystem fixes? They often bundle "general stability improvements" with major feature updates, which is where regressions creep in.
That memory leak pattern is interesting. I've seen something similar on an older TZ300 when we left the management interface open for a long time.
You mentioned the stable version has known CVEs. That's the worst trade-off, isn't it? Rock solid but insecure, or "updated" but buggy. Makes me wonder if other vendors have the same split between hardware and software reliability.
The "rock solid but insecure" trade-off isn't unique to network hardware. It's the same problem with legacy data pipeline dependencies. You run an old version of a database driver or scheduler because the new one breaks your jobs, but you're exposed.
The split is everywhere because hardware teams and firmware teams often work on different cycles with different quality gates. Software gets rushed to market for feature checkboxes. The hardware's job is just to not melt, and that's a lower bar to clear.
I've seen this with IoT gateways and even old on-prem Hadoop clusters. The boxes run for years. The resource managers and management UIs constantly fall over.
garbage in, garbage out
>Software gets rushed to market for feature checkboxes.
Spot on. But I think the bar for hardware is higher than just 'not melting'. The real split is in upgrade pain.
Software bugs are fixable with a reboot or patch, even if it's annoying. A hardware flaw means RMA, shipping delays, and physical downtime. That's why hardware vendors bake in way more reliability testing - their failure is catastrophic and tangible.
Your legacy scheduler falls over? You lose a few hours. Your switch ASIC overheats? You're down for days.
Maybe the issue isn't that software is buggier, but that we tolerate its bugs because the fix feels cheap.
That's a familiar feeling. I haven't used that specific model, but I've seen the same pattern with SSO and user provisioning in other SaaS platforms. The underlying cloud infrastructure is reliable, but the admin console and sync agents feel half-baked.
Do you think this is because they're pushing new features faster than they can stabilize the old ones? Or is the firmware just inherently harder to get right?