Skynet Skynet and Network Speed Test slow-downs ..

[SpasilliumNexus]

Sharing my results with Skynet enabled and disabled. My speeds are faster when it's disabled.

RT-AX88U, completely clean reset and set up from scratch on 386.1.
Verizon Gigabit FiOS.
240GB Kingston SSD in a UGREEN 2.5" enclosure, with SWAP enabled.

EDIT: AiProtection is on, if that makes any difference.

 

[roboots21]

Personally I feel so much more comfortable online with Skynet I would gladly accept a 50% speed cut for the protection, but YMMV.

Edit: I've also never paid for service faster than 200Mbit.

Sharing my results with Skynet enabled and disabled. My speeds are faster when it's disabled.

RT-AX88U, completely clean reset and set up from scratch on 386.1.
Verizon Gigabit FiOS.
240GB Kingston SSD in a UGREEN 2.5" enclosure, with SWAP enabled.

EDIT: AiProtection is on, if that makes any difference.

Albeit different speeds, this is the behavior I am seeing as well.

 

[capncybo]

Skynet is NOT actually slowing down your downloads at all.
Its slowing down The Java Based Web Test that is run via the Router httpd.

If you have a fairly recent Phone with decent 5G wifi...
Try & Test your Speeds using an Ookla app on the phone instead.
Or... If you have a fast computer with a 5G (or better) connection such as wired.

Just Test using that.

Go ahead & turn Skynet ON & OFF...
You won't see those massive discrepancies of speeds.

So, what you are witnessing is some of the display limitations of your router.
It's optimized to route data packets, not for rapid Java GUI type applications.
As Network data speeds get closer & closer to CPU speeds...
The CPU can't possibly keep up.

Relax, Turn Skynet Back-On & Enjoy ;-)

 

[capncybo]

I've actually tried, and there is no real difference. I still get the slow down in speed every time I enable Skynet.

I've even plugged a PC in the router and ran the test from the PC. I get the same slow down.

This only happens with SKYNET.

Sorry, didn't see any pictures of speed tests from anything but the routers web based speed test. So you are saying you can't see those upper speeds on your pc plugged into ethernet. Are you 100% sure you get those upper speeds on your ethernet pc. I ask because on some budget priced laptops... The bloody LAN card is only 100mb but the WiFi will allow faster throughput.

 

[capncybo]

Yeah, I don't see the speeds either if I plug my Laptop directly to my router using a CAT 6 cable. I do see the full speeds (430+Mbs) with the same laptop connected to the router with skynet disabled however. This means that it has nothing to do with Java or anything. The only varible is skynet.

I love skynet, but a speed decrease of this much is too much.

I had a good sleep & woke up with a clear head. Things are only ever as fast as their slowest point.

I'm thinking the limiting factor is the actual (Read/Write) I/O of the USB-Controller.
I quickly looked up your router via ASUS website...

"RT-AC5300's top-of-the-line 1.4 GHz dual-core processor brings its computational capability to the next level. USB data transfers enjoy up to over 100 MB/s speed"

If you are accessing Skynet & Swap memory on something that is (Around 100 MB/s @BEST)
How can you possibly expect the CPU to read information, perform some routing... & update a WEB-GUI all at speeds that greatly exceed the maximum throughput of the built-in USB-Controller?

I have a RT-AC68U & its great for an old device but it cannot Read/Write to a USB-drive at even 100 MB/s.
These routers are not NAS-Drives with super quick USB-transfer.

I kinda fell for this when I first saw the USB 3 port on my router..
But just because USB-3 supports a maximum of up to 4,800 Mbps = 600 MB/s
Doesn't mean it can actually do so...
In reality it's more like 480 Mbps = 60 MB/s

If reading & accessing skynet at speeds like 60 MB/s...
EVERYTHING Better be in the routers cached memory or else... "BAM" accessing slower swap-memory at approx 60 MB/s

And I think SpasilliumNexus
was actually talking about 900+MB/s... while you were talking around 400+MB/s

But either-way,
The routers actual USB-Transfer speeds seem somewhat inadequate for the task.
Especially if its being used as swap-memory.

 

[AntonK]

S

Thanks for the follow-up. I am a little disappointed by the specs of the USB controller for my router, but it makes sense if that is the case. However, one thing I noticed, and not sure if it makes a difference, my swap file seems to never be used. The GUI shows that I have a swap file, but the router is not storing any files in it.

Is there a way to force SKYNET to use only the router cache? Maybe that would fix it....

Anyways, just a thought, but it seems that, like you said, it might be my USB controller :-(

Same here regarding swap file usage. Regardless of the exhortations to make sure you've got a 2GB swap file, I'm yet to use 1 byte of it.

 

[aex.perez]

AX88U Merlin 386.1, 384.19 and all the beta's in between, 1Gb symmetric fiber, a few scripts including Skynet. In all the testing scenarios at the router, the Windows clients with the app, ~940Mbps down and up. Consistently, which is what matters to me.
I did some extensive testing, controlled as one could control over the internet. Using the same target, 15 runs, averaging out the results. TrendMicro on, any related feature enabled that requires acceptance of the privacy agreement. Cuts my results almost in half with uploads faster than downloads by ~100Mbps. Disable all the TrendMicro related features off, without Skynet, without withdrawing from the TrendMicro Agreement, results matched the Windows Client. With Skynet on, TrendMicro related features disabled but without privacy withdrawn, monitoring only inbound, results matched Windows client. But after about 18 hrs, something triggered TrendMicro and saw my results halved. Only after withdrawing from TrendMicro agreement did my results recover.

Final results with Skynet monitoring inbound and outbound, withdrawn from TrendMicro Privacy agreement, ~920Mbps download, ~650Mbps upload and that's using the same target, and its consistent. Still the Windows client ~940 up/down, and that's where I sit today.

I emphasize, in all testing scenarios, no variations in the client results. I even enabled priority for the Router on the AT&T Modem, no change. I suspect if I was using the modem for anything else but a passthrough for the router I might have seen a difference.

I use spdMerlin to monitor my speeds (uses same Speedtest/Ookla as ASUS) and as long as I am consistent with my results I'm happy regardless of the number result. Ping, Jitter, Packet Loss are more important and impactful to me. If I see a large variation in the trend lines/graph regarding speed results then I'll worry and start checking the devices and other Speedtest targets to see if I get closer to what my ISP provisioned. I'm not hung up with the Router Speedtest results matching what the ISP provisioned.

One thing to look at is router CPU utilization during a Speedtest at the Router vs. (in my case) the windows PC. The impact at the Router is much more pronounced with the Router runs vs the PC runs, as there is a dependency on the CPU of the device running the Speedtest for the Speedtest itself and it will impact your results. Depending on your Router's CPU, features enabled, TrendMicro, what its running (scripts), your selected Speedtest server, and the Internet/carriers your results and mileage will vary.

 

[archiel]

Looking at this thread

• some users are noting a performance hit as a result of using Skynet with a fast connection (sadly not a problem for me 80/20 )
• it is suggested that the data transfer speed between the router and the usb device may be the issue
• some users have observed that the swap file does not appear be used, or only used minimally (mine shows 17.14 Mb used)

How can you see/test the data transfer rate between the router and the attached usb?
How can you see / what traffic is using the swap file?

 

[capncybo]

AX88U Merlin 386.1, 384.19 and all the beta's in between, 1Gb symmetric fiber, a few scripts including Skynet. In all the testing scenarios at the router, the Windows clients with the app, ~940Mbps down and up. Consistently, which is what matters to me.
I did some extensive testing, controlled as one could control over the internet. Using the same target, 15 runs, averaging out the results. TrendMicro on, any related feature enabled that requires acceptance of the privacy agreement. Cuts my results almost in half with uploads faster than downloads by ~100Mbps. Disable all the TrendMicro related features off, without Skynet, without withdrawing from the TrendMicro Agreement, results matched the Windows Client. With Skynet on, TrendMicro related features disabled but without privacy withdrawn, monitoring only inbound, results matched Windows client. But after about 18 hrs, something triggered TrendMicro and saw my results halved. Only after withdrawing from TrendMicro agreement did my results recover.

Final results with Skynet monitoring inbound and outbound, withdrawn from TrendMicro Privacy agreement, ~920Mbps download, ~650Mbps upload and that's using the same target, and its consistent. Still the Windows client ~940 up/down, and that's where I sit today.

I emphasize, in all testing scenarios, no variations in the client results. I even enabled priority for the Router on the AT&T Modem, no change. I suspect if I was using the modem for anything else but a passthrough for the router I might have seen a difference.

I use spdMerlin to monitor my speeds (uses same Speedtest/Ookla as ASUS) and as long as I am consistent with my results I'm happy regardless of the number result. Ping, Jitter, Packet Loss are more important and impactful to me. If I see a large variation in the trend lines/graph regarding speed results then I'll worry and start checking the devices and other Speedtest targets to see if I get closer to what my ISP provisioned. I'm not hung up with the Router Speedtest results matching what the ISP provisioned.

One thing to look at is router CPU utilization during a Speedtest at the Router vs. (in my case) the windows PC. The impact at the Router is much more pronounced with the Router runs vs the PC runs, as there is a dependency on the CPU of the device running the Speedtest for the Speedtest itself and it will impact your results. Depending on your Router's CPU, features enabled, TrendMicro, what its running (scripts), your selected Speedtest server, and the Internet/carriers your results and mileage will vary.

This is truly Excellent troubleshooting & thank you very-much for sharing your results.
Your Experiences are almost identical to my-own recent findings.
Except that my ISP package limits me to 300/15 & My router is a slower RT-AC68U.
I had recently noticed the correlation between enabling & disabling Trend Micro & opting-out of the privacy agreement.

+ As Others have mentioned (& looking closely at it), Swap file "usage" does seem rather minimal.

@archiel - I think the easiest way to view realtime swap-file use is... "htop".​

Install it via: opkg install htop
& run it via : htop
Or, simply type: free -m
for a snapshot of swap & other memory usage.

@TheMorpN - "However, one thing I noticed, and not sure if it makes a difference, my swap file seems to never be used. The GUI shows that I have a swap file, but the router is not storing any files in it.
Is there a way to force SKYNET to use only the router cache? Maybe that would fix it...."

That got me thinking to adjust swappiness & I tried (but I didn't see a notable difference).
via:
sysctl vm.swappiness=60
sysctl vm.swappiness=10
sysctl vm.swappiness=90
sysctl vm.swappiness=0
sysctl vm.swappiness=100

However what I did learn (& myself included) is that swap-file usage is GREATLY misunderstood.
See Busting Myths about Swapiness

What Is Swappiness on Linux? (and How to Change It)

Swappiness in Linux has nothing to do with how much RAM is used before swapping starts. We tell you what it really is.

www.howtogeek.com

But regardless... if the swap-file was responsible I still thought the adjustments above should have had a slight impact or made a some-difference, hhmmmmm ???

Anyways in my case...
I'm only using a:
SWAP File; /tmp/mnt/sda1/myswap_1G.swp (1.0G)
total used free
Swap: 1046524 54156 992368

And much like archiel when I have skynet enabled & run tests on a 5g Samsung S8 I can on-average hit the MAX provided throughput of 330 down & 15 up.
But NOT on the Asus RT-AC68U web GUI Ookla Test...
With Skynet & Trend Micro Enabled... Like others its half.
But as I tried to explain above... those numbers are NOT accurate.
What matters is... that the other fast cpu devices in the family &
their speedy cards can hit the MAXIMUM ISP speeds.
And they can.

I don't use our family-laptop's wired ethernet connection plugged into the router because its a crappy 100Mb card & the built-in WiFi is faster... WTF (I know).
Anyways, the teenager has a brand new playstation-5,
I'm curious what numbers it can get, but my bet is +300 no problem.

Anyways even if the swap-file is not to blame...
when looking at the skynet firewall:

FW Version; 386.1_0 (Jan 30 2021) (2.6.36.4brcmarm)
Install Dir; /tmp/mnt/sda1/skynet (26.7G / 31.0G Space Available)
SWAP File; /tmp/mnt/sda1/myswap_1G.swp (1.0G)

The software was installed to my usb-drive & swap is on the usb-drive
So the router & CPU obviously have to access this information periodically.
And the transfer speed is non-steller.

Yes I could UN-install trend micro by opting out, But I also noticed that under skynets [11] settings...
[7] --> Import AiProtect Data | [Enabled]

Perhaps this setting could be Disabled but my other confusion is...
If you opt-out of Trend Micro... Won't Skynet loose access to Trend Micro AiProtect Data?
Point being...
With BOTH enabled, isn't it redundant?
The Final annoyance (I've yet to verify) is...
If you Do NOT Opt-Out from Trend Micro & simply disable via the WebGUI.
Ive read Trend Micro Turns itself BACK-ON.

Thinking about it, Perhaps Its actually the Skynet Setting above which continues using AiProtect.

 

[JaimeZX]

All Skynet's Ban AiProtect feature does is add any AiProtect-flagged IPs to your local blacklist. Skynet doesn't have any magic access to TrendMicro's servers.

 

[capncybo]

All Skynet's Ban AiProtect feature does is add any AiProtect-flagged IPs to your local blacklist. Skynet doesn't have any magic access to TrendMicro's servers.

Thanks, I suppose these additions are accumulative/additive? If so... theoretically after they've been added. Even after someone Opts-Out from Trend Micro...
- the Banned IP-addresses will still remain (And That's Good).

Of course the irony in...
Turning-off all security layers & Opting-out of almost every option the router comes with such as QOS etc kinda turns these things into a fairly standard router or at its most basic... a switch.
And simply doing it for a more accurate ROUTER/Web-based Ookla speed test...
Makes Zero Sense.

 

[JaimeZX]

Sure, the previously-banned IPs will stay in the Skynet blacklist. That'll have diminishing benefits though, since bad guys tend to move around.

I will say I don't think just turning off TrendMicro services after turning them on does anything unless you do the TM Opt-Out.

Personally I stick with Asus for the security, so I'll gladly run TrendMicro + Skynet + Diversion and take any hit, even 50% if that's the price. But as I said earlier, none of the places I have these routers has over 200Mbit service so I haven't noticed anything.

 

[Gar]

The idiot light on my USB stick flickers often in normal use, which I assume is read/write activity on the AC56. I cannot test AI Protection but the speeds I measure are nearly identical no matter which scripts are running with 100/100 fiber.

 

[L&LD]

Can't have one without the other.

 

[capncybo]

I'm a little confused. If I "Withdraw" from the AiProtection, Traffic Analyzer, Apps Analyzer, etc, from the Administration/Privacy tab, does that mean that I would lose all info in Traffic analyzer etc, or does it mean that my Router will NOT send any data to ASUS about AiProtection, etc?

I do still want to keep the features.

Actually you don't sound confused, you've just become realistic & more educated. The speeds we are talking about here are approching what used to be ISP-provider/institutional...
And for example the absolute Max USB-3 Transfer speed = 600MB/s... But we wanna GO FASTER !!!
Gonna need us a Newer More Powerful Router it would seem.

EDIT: But IMO you are right to feel somewhat deceived when a company advertises USB-3 connectivity & it doesn't even deliver the Maximum USB-2 throughput.

 

[capncybo]

My Curiosity got the better of me, so... to mimic the router accessing an Entware based installation of Skynet...
-I installed iperf3 via: opkg install iperf3
-Then I ran the entware installed iperf3 in server mode set for MB/s as a daemon: iperf3 -s -f M -D
-Next I ran the entware installed iperf3 as a client while set for MB/s via: iperf3 -c 192.168.1.1 -f M

The output showed a Bitrate transfer averaging out @ just under: 90 MBytes/sec
There's my "real-world" USB-Drive speed... I would certainly say that is a bit of a BOTTLENECK!!!
Especially when dealing with network speeds far, far, faster.

 

[JaimeZX]

But you guys aren't reading and writing all your traffic through USB; Skynet is resident in memory and just stored on the external drive. With as little as the swap file gets used already, I'm not clear what about the USB speed you think is a bottleneck.

 

[SomeWhereOverTheRainBow]

But you guys aren't reading and writing all your traffic through USB; Skynet is resident in memory and just stored on the external drive. With as little as the swap file gets used already, I'm not clear what about the USB speed you think is a bottleneck.

I concur, others might argue that iperf3 application is being limited by the usb it is installed on.

 

[capncybo]

Yeah, poor choice of wording on my part with "There's my "real-world" USB-Drive speed..."
But what I was perhaps failing to point out was... Wouldn't this be an example of the ABSOLUTE Fastest any Entware Loaded apps could possibly I/O as it's just a loopback onto itself?
Just as @SomeWhereOverTheRainBow had pointed out above.
And I probably need another Linux computer to test, but if I were to actually transfer a file onto that usb-drive the actual transfer rate will probably be about half (Approx 45 MBytes/sec) at best.

Regardless, If I'm understanding your points correctly...
The memory-speed & CPU limitations are perhaps the REAL limiting factors.
It's my understanding that "Except" when Skynet is doing the updates... most of the required code for packet filtering should already be loaded into the router's resident memory.

For myself... understanding Why Asuswrt-Merlin's Built-In Ookla Speed test is less accurate(or too slow) when Skynet & TrendMicro is Enabled...
Has become more of a Mental Exercise or a Thought Experiment.

At my speeds of around 300down/15up...
Some of my devices slightly exceed those numbers with BOTH Skynet & TrendMicro Enabled when speed test is run on the "client device".
So I'm just trying to better understand...
What actually prevents the built-in Speed test & the Entware/amtm spdMerlin tests from being able to do so
And why they can obtain the higher numbers if BOTH Skynet & TrendMicro are Disabled.

As I said in a post above, I do understand...
Processes are always limited by their slowest point.

 

[SomeWhereOverTheRainBow]

Yeah, poor choice of wording on my part with "There's my "real-world" USB-Drive speed..."
But what I was perhaps failing to point out was... Wouldn't this be an example of the ABSOLUTE Fastest any Entware Loaded apps could possibly I/O as it's just a loopback onto itself?
Just as @SomeWhereOverTheRainBow had pointed out above.
And I probably need another Linux computer to test, but if I were to actually transfer a file onto that usb-drive the actual transfer rate will probably be about half (Approx 45 MBytes/sec) at best.

Regardless, If I'm understanding your points correctly...
The memory-speed & CPU limitations are perhaps the REAL limiting factors.
It's my understanding that "Except" when Skynet is doing the updates... most of the required code for packet filtering should already be loaded into the router's resident memory.

For myself... understanding Why Asuswrt-Merlin's Built-In Ookla Speed test is less accurate(or too slow) when Skynet & TrendMicro are Enabled...
Has become more of a Mental Exercise or a Thought Experiment.

At my speeds of around 300down/15up...
Some of my devices slightly exceed those numbers with BOTH Skynet & TrendMicro Enabled when speed test is run on the "client device".
So I'm just trying to better understand...
What actually prevents the built-in Speed test & the Entware/amtm spdMerlin tests from being able to do so
And why they can obtain the higher numbers if BOTH Skynet & TrendMicro are Disabled.

As I said in a post above, I do understand...
Processes are always limited by their slowest point.

From my understanding, the biggest limiting factor is TrendMicro because it uses the CPU to place itself in-front of traffic as each packet gets inspected for threat level protection. Users of skynet have previously tested disabling AIprotection and these users also concluded they had to withdraw from these disclaimers as well. Skynet will continue to work as it does without assistance of AI-Protect.

Following a reboot, your speeds should be noticeably faster.