I finally got around to setting up a DMZ subnet on the EPIA PD-10000 1.0Ghz C3 board w/1GB ram and ran some network performance tests through the ASG.
Setup:
Eth0: 192.168.0.xx Port 1 on Intel Dual Port Pro/1000MT card (internal network)
Eth1: 192.168.1.xx Port 2 on Intel Dual Port Pro/1000MT card (DMZ w/web servers).
Each subnet above is connected to it's own Netgear gigabit switch (the 0.x internal subnet is on a 8-port GS108 gigabit switch and the 1.x DMZ subnet is on a 5-port GS105 gigabit switch).
Eth2: External interface to cable modem 100BaseT connection using the on-board VIA VT6105 [Rhine-III] NIC.
Feature usage: ASG 6.303, 370MB of ram in use (of 1024MB)
- 8 packet filters
- DNS (on Internal & DMZ)
- DHCP (on Internal & DMZ)
- NO other proxy services
- 2 constant IPSec VPN tunnels to other locations
- Intrusion Protection enabled.
Here's where things get interesting:
From the external interface inward, everything is gigabit ethernet.
I have two systems on my "Internal" network that can transfer large ISO files at about 18MB/s back & forth.
These speeds & connections, while not that great for gigabit, do NOT go through the ASG.
However, when moving one of the systems onto the DMZ, which forces ALL traffic between the two machines to go through the Intel Dual Port gigabit NIC's in ASG, my transfers dropped to only 2MB/s as configured above.
This was a lot slower than I expected.
Disabling Intrusion Protection bumped the speeds up to a respectable 12MB/s through the ASG gateway.
I assume the Intrusion Protection service requires quit a bit of CPU horsepower to keep the throughput through ASG at a respectable level.
My current ASG setup seems more than adequate with IPS enabled for most decent sized operations and networks.
2MB/s sustained equates to about a 16mbps pipeline.
My cable modem service provides an 8mbps connection, which even at full saturation, would still leave about 50% capacity available on ASG for growth and bandwidth expansion.
However, I'm not sure how resource utilization and efficiency will change with the upcoming ASG v7.
It would be interesting to see how the Via C7 1.5ghz CPU performs in testing similar to what I did above. Hindsight 20/20, I wish I would have done this test before I deployed the Via EN-15000 on my relative's network.
This thread was automatically locked due to age.