I could not find any figures about real throughput (MB/s) and latency.
Are These available somewhere?
Right… someone could post real-life measurements, but I’m fairly sure it won’t be noticeable for over 99% use cases.
Tierisch Schild! Choose your favorite animal to protect you:
- or for your usual “home” network,
- up to the most fearsome dragon-based fire wall (which could sometimes eat also your normal traffic).
I have posted some easy measuremente on reddit some time ago.
omnia is able to transfer/NAT 1 gbps (maximum on ethernet, I achieved some 932mbit/s) stream.
with some expensive applications the performance can be lower.
I measured 550/550 Mbps SQM traffic shaping with bi directionally saturating traffic (not PPPoE, but NAT and default firewall) some time ago. Pretty respectable. Unidirectional traffic shaping worked up to a full Gbps, but that is a rather dicy proposition on a symmetric link.
EDIT: Forgot to explicitly mention that this was on a turris omnia under TOS4 or 5 (don’t remember exactly)
Guessing Shield’s performance based on Omnia’s… seems a bit tricky.
good point.
Omnia has 2-core Armada 385 ARMv7 @1.6ghz, reporting 2x 1600.00 BogoMIPS
Shield has 2-core Armada 3720 ARMv8 @1,0 GHz
I can only guess that the shield’s CPU reports 2x 1000 mips, so the performance might be over 60%.
…that should still be able to NAT full 1 gpbs with a few % CPU load. but the performance with huge number of packets and monitoring functionality will be lower.
The question is kind-of answered here (the first one but in CZ): https://www.root.cz/online-rozhovory/ptejte-se-michala-hruseckeho-vyvojare-projektu-turris/?pi=2
EN: both Omnia and MOX can handle 1 Gbps passing through in default setting (with NAT, measured by iperf3). As noted, Shield uses the HW of MOX. I believe that with Sentinel (that dynamic firewall) this throughput should be unaffected; it’s unclear to me if it was included in those measurements.
@mbehun I see you were responding on support. Could you please share some numbers here as well? Thanks!