Firmware - 6.2.0 2


#1

Hi,
Can you please tell me how the following new features work exactly, and any recommended threshold? I think they will be helpful to deal with some packet loss/latency issues we are having with Citrix but documentation is not really available anywhere yet.

  1. Cut-off latency2. Packet lost pull-back time
  2. Packet Redundancy

Thanks,
Derek


#2

Hi Derek,

Are you referring to SpeedFusion Bonding with 3 items above?


#3

Hi TK,

Under Speed fusion VPN setup, see images below:




#4

Hi Derek, you asked the questions just in time. We have prepared a technical notes describing these SpeedFusion features and posted to the forum:

Technical notes about advanced SpeedFusion features in firmware 6.2.0
https://forum.peplink.com/threads/3740-Technical-notes-about-advanced-SpeedFusion-features-in-firmware-6-2-0

Hope this can help answering your questions, and if you are still using 6.2 open-beta-1, you many want to upgrade it to beta 2, as we just released it today:
https://forum.peplink.com/threads/3736-Firmware-6-2-Open-Beta-Release-2


#5

Hi Steve,
Just wanted to let you know we have been using 6.2 beta 2 since 26th of September without issue and are very happy with the performance and reliability.
We were trialing the product for 4G bonding and were finding the 6.1.2 builds to be pretty poor, so much so it was jeopardizing the purchase. There seems to be some vast improvements in 6.2 beta2 and I have found the cut-off latency and packet redundancy options have helped immensely.

Is there any news of when of when this build might be finalised?


#6

Thank you for the good news, Derek, firmware 6.2 GA will be available soon (I expect RC version will release in coming weeks, and GA after that), but we don’t have an exact release date yet. Please stay tuned.

Would you mind share with us how the “cut-off latency” and “packet redundancy” help in your environment? How much does it improve and what’s the parameters value in your configuration? Thank you :slight_smile:


#7

Sure.
We use Citrix XenApp 7.6 and Avaya VoIP phones (H.323 & RTP) almost exclusively on all of our remote sites. We are a Construction company so often have issues with lack of infrastructure (lead-in & pits) for fixed services on site. We also need to be able to setup camp extremely quickly. So our aim was to make use of more 3G/4G services.

Unfortunately, cellular services have good and bad days from a latency and reliability perspective. Even when you have a huge antenna pointed at a tower and lock frequencies.

The problem I was seeing with older version of the firmware was that while it was quite good for large data transfers, it was inconsistent for session protocols. There was consistent freezing on Citrix due links spiking in latency or dropping packets. VoIP also had issues as well. The only way I could get it to work fairly constantly was in fail-over which is not what I want.

Across our initial deployment (3 sites) we now have HD4 devices with reasonable LTE reception (Telstra - Australia) using the included antennas. The settings we are using on these sites are “packet redundancy” = “Repeat Once” and “Cut-off Latency” = 250. Since using these settings, particularly packet redundancy, there was instant and hugely noticeable change. So far there has been no complaints from sites and only positive feedback from the people I have asked about it.

Time will tell but so far thing are looking really good. I’m interested to see what happens when we hit our first 3G only site or really poor reception. I imagine we will need to tweak these settings a bit.


#8

I will echo these comments on the new cellular speedfusion features, what I’ve seen is working well

We use exclusive cellular connections regularly in static and on-the-move too and I’d suggest trying a good MIMO antenna with some gain
The supplied antennas are fine for irregular cellular traffic but to get the best out of them a good antenna is recommended by me at least
I can 3G or 4G sometimes when it appears theres nothing there to other equipment


#9

Hi Derek, appreciate for your help. Did you try “Repeat twice” as well? Do you see any differences? Actually I have a small concern about the extra bandwidth required by this feature, what’s the average required bandwidth on your site? and how much bandwidth can you get from the LTE connection? You talked about Telstra coverage, so are you using 4 x Telstra SIM on HD4?

Hi Michael, do you mean you have tried cut-off latency and it works well? or do you mean the other new SpeedFusion features? :slight_smile:


#10

Yes the cut-off latency seems to work for me and the lost packet feature too

I’ve not sat monitoring it during heavy use for a couple of weeks but I’ve had no trouble which is the point really isn’t it


#11

Hi Steve,
Sorry I never got back to you as I simply didn’t realise you replied. Yes, I tried Repeat Twice, it was reliable but “felt” slower (Citrix). Possibly the extra traffic or processing was causing extra latency?

Frankly I don’t really care too much about bandwidth, as my use case is for VoIP/Citrix traffic in locations where its hard to get fixed services. My aim is for reliability and lowest possible latency. If you had an option that mirrored all the traffic over two cards (or more) and discarded the slowest packet to arrive I would use that despite the cost. Ideally I like to get about 7 mbps down and 1.5 mbps up on each site.

I’ve have a bunch of Dual SIM HD2 and HD4 devices. I don’t really see any improvement from using additional cards. In fact it tends to make latency/jitter worse. If I add more cards I tend to put them in standby.

All that being said, has this option been changed in Beta3(other than the name)? Unfortunately, it doesn’t seem to be working as well for me anymore.


#12

Hi DerekH,

Sorry for the delayed reply, Derek, as you have aware about the name change, we are also keep improving the algorithm in the backend, which we call it “WAN Smoothing” now, the purpose of this feature is to reduce the impact of packet loss and get the lowest possible latency, at the expense of extra bandwidth consumption. This is exactly what you have described in your post. What is your test result with / without this feature enabled (using repeat once as in your firmware)?


#13

Hi Steve,
I assumed this is how it worked but never tried any captures to figure it out. However, the reason I queried this theory was due to my observations in B3. This function doesn’t seem to be working as well as it did in B2. If I use one cellular card on its own (with the setting off) things tend to be much smoother than if I use two and enable these settings. There is freezing in Citrix sessions and breakups in VoIP calls.

This was not the case in B2, if I enabled “Single” I would see a major positive change. “Double” was also a noticeable change. Obviously, all of this is a perceived observation and based on a network with no guarantee so take it as you will.

The other observation that I have is that the throughput just drops through the floor. Correct me if I’m wrong but if we’re mirroring traffic then theoretically we should be able to get speed equal to one card minus some overhead?

The results of transferring a file over the VPN with SMB:

  • One Cellular modem no latency reduction = 12mbps (bottleneck being the remote end expected)
  • Two cellular modems with no latency reduction = 12mbps (bottleneck being the remote end as expected)
  • Two cellular modems with “Normal” latency reduction = 2mbps

I have a spare HD2 device and a HD4 device with active LTE services not in production for the rest of this week if anyone is interested in taking a look. The test units are in fixed locations with good reception.

Also, would you be able to tell me what happens if this setting is on and there is only one connection, is the packet transmitted more than once on the one connection or is it overridden? In the case there is only one active connection on one end and two on the other does it work asymmetrically?

Thanks,
Derek


#14

Hi Derek,

Sure we’d like to take a look, please enable Remote Assistance (http://www.peplink.com/knowledgebase/how-to-enable-the-remote-assistance-service/) and open a support ticket here:
https://cs.peplink.com/contact/support/

Wan Smoothing only have effect when there is more than one WAN on either both or one of the peers. So if the remote SpeedFusion peer has only one WAN, you’ll need 2 WAN connection on local for WAN Smoothing to be effective. For any 1 to 1 WAN SpeedFusion, WAN Smoothing won’t kick in at all. Also please be reminded that WAN Smoothing is an option for transmit direction, to have effect on both Tx/Rx direction, you’ll need to enable WAN Smoothing on both side.


#15

I’ll log a ticket now.

This is odd, if I enable two connections on the remote end with normal mode smoothing I see the exact same download speed from a service behind this side as if I do it with one connection (2mbps). Turn it off and everything picks up again.

Derek


#16

Dear Guys,

I’m not satisfy with Bandwidth Control on this beta version. No network admin wants to limit all computers inside the network. it should be configurable per IP or MAC Address. also in Captive portal we need a white list to grant Internet access bypassing captive portal. these two settings are very general and useless in most network environments. they should allow network admin more granular settings.

I hope you consider this for the final release.

Best,

Hootan


#17

Just posting back in case anyone is interested in the outcome.

With Steve’s help we were able to identify that our primary connection used on our core balance router was dropping an excessive amount of UDP traffic but having no issues with TCP. The outcome was that Telstra IPVAS (Juniper Firewall) service has UDP flood protection enabled at the default rate of 1000 pps. This function is not host specific and either needs to be increase or removed for from the WAN interface of your virtual firewall (IPVAS). If you are in the same boat I would suggest trying increasing this threshold to at least 5000pps or higher as the PepVPN UDP traffic is very aggressive.

As a result of this I expect the change I have seen between 6.2.0 b2 and 6.2.0 b3 using the WAN smoothing function was as a result of an increase in the number of PepVPNs we now have online, which has caused us to hit the threshold. We are still testing but things are looking much better now.

Thanks for your help Steve,
Derek