Latency on WildBlue

[Valstaag]

Before I talk about latency, let me tall you that my speeds are just fine. Even during peak hours, my speeds are almost always near their max. This post is about an even more complex issue, latency. Specifically latency of UDP traffic. I welcome all responses on this topic, so long as they are articulated intelligently. Responses like : 'YEAH!, WB sucks' or 'You don't know what you're talking about, WB is GREAT!' will be disregarded.

As some of you know, and have posted on testmy.net and DSLreports.com, the latency of the WB system has approximately DOUBLED ACROSS THE BOARD. Normal operation for a properly running WB system in terms of latency used to be 600-700 Milliseconds between the modem and the NOC (the best way I know of to check this is to run a constant ping on the ip address of your default gateway (Start->Run->Type CMD to load a DOS prompt, type ipconfig to find out your default gateway, type 'ping xxx.xxx.xxx.xxx -t' where xxx.xxx.xxx.xxx is your default gateway as identified by ipconfig)) After a suitable sample has been obtained (a half hour is a pretty good sample(make sure that you are not using the connection during testing!) press control c to quit the ping and get the stats. They will look something like this:

Ping statistics for 75.104.36.1:

    Packets: Sent = 1175, Received = 1167, Lost = 8 (0% loss),

Approximate round trip times in milliseconds:

    Minimum = 1080ms, Maximum = 3766ms, Average = 1226ms

(FYI, these stats were taken after primetime on a properly installed dish on Cheyenne beam 24 on a new build of XP Sp2 which has all MS security patches, is free of viruses and adware as certified by AVG, Ad-Aware, and Spybot S&D (all industry accepted AV and AA apps), is well beyond the system requirements of WB, there is no third part hardware in line, and all unnecessary programs and processes have been stopped. These are in my experience as good as you can now get through WB at any time, on any gateway after aforementioned 'update/upgrade').

Averages seem to go up 200-800 ms during peak usage hours, and an additional 200-800 ms when pinging sites on 'the internet' during peak hours(google.com,yahoo.com,WOW gameserver, etc.('internet' connectivity has more to do with WB's ISP (ATT) than the WB equipment, however it is still WB's

responsibility to ensure the quality of said access to the internet, as WB is reselling that access and representing it as their own)). 

As far as latency is concerned, WB has the following statements on it's web site:

From WB.com : http://wildblue.com/aboutWildblue/qaa.jsp :

30. Will I be able to use a virtual private network (VPN) with the WildBlue service? Yes. WildBlue recommends the use of SSL-based VPNs. They work well on the WildBlue network because they work efficiently over a satellite connection. IP Sec-based VPNs may work over the WildBlue network, but will likely be much slower (and perhaps not useable) due to the effects of latency.

31. What is the impact of latency? Can I play real-time Internet games or make Internet phone calls on your service? The WildBlue system is engineered to help offset the impact of latency, which is the delay caused by sending signals from the earth to the satellite and back again. However, there is a delay of about a half second as the signal travels up to the satellite, back down to the gateway, up to the satellite and back down to your modem. For most applications this latency does not affect performance, however, there are some applications like voice over IP (telephone service delivered over the Internet, also known as VOIP), or real-time interactive gaming, where latency will have a noticeable effect on performance over the WildBlue network, as it would on any satellite-delivered service.

Now, let's take a look at the first statement of note in qaa 31: "The WildBlue system is engineered to help offset the impact of latency." This claim (especially in light of recent changes) seems laughable, but it does have a small point of merit, in that WB uses the KA band, which uses a higher frequency than the more commonly used KU band, and higher frequencies DO tend to travel faster. Whether this has a noticeable effect in practice (or even theoretically) is debateable, however. 

The second statement of note is the one that sets expectations in terms of latency: 'However, there is a delay of about a half second as the signal travels up to the satellite, back down to the gateway, up to the satellite and back down to your modem.'

For those uninitiated with si prefixes, (milli=thousandth), 500ms = five hundred thousandths of a second = .5 or 1/2 second.

That statement, as available now on the WB corporate web site is setting customer expectations in terms of latency at 500 ms. This is unrealistic in the best of conditions on the WB satellite system. As documented by customers' posts on testmy.net and DSLreports.com; before the update, on the best beam, on a well installed dish with no CPU issues, latencies of 600-700 ms were the rule as measured between the modem and the satellite. this translates to 700-800 ms between the customers CPU and the internet (assuming good NOC communication) especially in the obvious context of gaming (which usually uses UDP protocol, which by default has a lower network priority, and when subjected to network traffic control techniques such as traffic shaping and TCP synchronization oftentimes becomes unusable in terms of timeouts and latency during times of high traffic.(one good paper on this is: "Characteristics of UDP Packet Loss: Effect of TCP Traffic" http://www.isoc.org/inet97/proceedings/F3/F3_1.HTM). Now, add on the obviously purposely implemented changes that came into effect in late November, and we are seeing a consistent, widespread UDP latency average of 1200-1400 ms (and higher, on the busier NOCs) as evidenced by ping tests. This is substantially greater than the 500ms (1/2 second) expectation set, and there is no wonder that even the people who were satisfied when the latency was 600-800 ms are now dissatisfied, and have even taken the trouble to change their ratings on DSLreports.

This isn't all, however. The statement continues, and definitely does not gain in credibility, nay even reality: 'For most applications this latency does not affect performance...'

The first part of this drawn out sentence is a flat out lie. The opposite is true: LATENCY AFFECTS MOST NETWORK APPLICATIONS. Even web browsing is noticeably affected by latency, contrary to WB's assertions. Whether WB likes it or not, The expectation of commercial network/internet communication is that it is as close to real-time as possible. Period. Even in this day and age, there is still a large percentage of users who cannot distinguish the difference between the performance of their network/internet and the performance of their computer, and when you add any wireless connection to that equation (much less something as intricate as satellite), you have a HUGE job of setting customers expectations appropriately. Even if you completely ignore QOI (quality of installation), service, and support issues, WB has still failed miserably in appropriately setting customers' expectations, even when the WB system was performing to the best of it's potential.

The rest of the statement continues as such: "...however, there are some applications like voice over IP (telephone service delivered over the Internet, also known as VIP), or real-time interactive gaming, where latency will have a noticeable effect on performance over the WildBlue network, as it would on any satellite-delivered service."

Well, at least they are admitting that latency DOES have some effect, although they are misconstruing it by saying it only affects gaming and VOIP, and they are setting the expectation in combination with previous sentences that the latency that will affect vpn, gaming, and voip is 500 ms. This is completely contrary to what support tells the customer. Support tells us that "ICMP packets are 'de-prioritized' in favor of TCP/IP traffic, and so pings and tracert tests are to be used for indicating connectivity only (basically saying that if you can ping something, it indicates connectivity, and if all packets are dropped, then there is no connectivity)". As stated previously, the first part of this sentence is correct. UDP packets are low priority for all network schema,configuration,etc. (From Wikipedia http://en.wikipedia.org/wiki/Best_effort_delivery : TCP verifies that all information transmitted is received fully on the other end. UDP does its best to deliver packets to the destination, but takes no steps to recover packets that are lost or misdirected. The more stable TCP protocol is often used to deliver data like web pages and email, while UDP is often used for media streaming or network gaming. ( see also (again): "Characteristics of UDP Packet Loss: Effect of TCP Traffic" http://www.isoc.org/inet97/proceedings/F3/F3_1.HTM )

Lastly, they are plainly stating that any limitations of the service are solely based upon the limitations of the technology by tagging on: "as it would on any satellite-delivered service."

My biggest concerns are as such:

1. WB does not discern between UDP and TCP traffic in their marketing info or on their website.

2. WB sets the customer expectation that 'Most applications' are not affected by latency in their marketing info and/or on their website, despite the fact that latency can even affect browser response to requested page loads.

3. WB sets the customer's expectations for latency on the few stated "applications" that are affected by stated latency (specifically IPsec VPN, Gaming, and VOIP) at 500ms.

4. WB did purposefully and without notice to it's customers change their network so that latency affecting UDP traffic approximately doubled.

5. WB COMPLETELY disregards the facts that there is great precedence of using PING and TRACERT commands to diagnose and measure performance on networks utilizing ALL mediums AND protocols, and it's results are even relevant to TCP since TCP travels over UDP.

6. WB is clearly saying one thing through it's sales (A half second latency, only affecting gaming, VPN, Voip, etc.) and something completely different through it's tech support (latency cannot be measured by ping tests, because ping uses UDP (even though those ping tests use the same UDP that gaming does), and therefore we give ourselves a blank check in regards to latency... as long as the TCP packet gets there someday, we have fulfilled our commitment. WB only TRULY supports browsing and email.) This is a glaring contradiction which betrays the consumer's trust, and may possibly be actionable as false advertising or even fraud.

Please do not think I am just a 'hater'. I think that the WB system has the potential of being the greatest wireless ISP around, and before the recent 'update/upgrade', it was meeting that potential for me, within the limitations of satellite communications. There was a small but noticeable hesitation when browsing, but nothing timed out. With a good system, even relatively bandwidth intensive on-line games and applications were playable, if barely. Now it is common to see 3+ second lag times during peak hours and beyond. Page load requests time out intermittently. This is on a well installed system on a good beam... I can only guess what it is like on some of the busier beams and/or with a sub-par installation.

For those who have contacted support and made an honest attempt to resolve their issues, I reccomend that you keep record of those conversations. If you have not contacted support, you should do so, the previously mentioned issue may be a result of a virus or ad/malware, a bad install, or faulty equipment. You need to rule those things out before addressing the network.

If you attempt to work things out with WB and fail, your options are somewhat limited. The few regulations affecting the telecommunications industry are outdated. There has been a movement around a concept called 'net-neutrality' ( see http://en.wikipedia.org/wiki/Network_neutrality ) which has at it's base the ideal that all internet traffic should be given equal priority. There are many arguments for and against such a thing, and it was somewhat moot with the previous Congress, but the Democrats are much more likely to place real regulation on businesses, so this may change. You might try sending your local Congressman a letter citing your concerns.

One Government entity that may be able to help is the FCC. The FCC is supposed to regulate all radio transmission, which satellite internet would surely fall under. It was the FCC that approved the use of the KA band for internet service. They themselves seem unsure, though.

From FCC.com:

http://www.fcc.gov/cgb/information_directory.html

The FCC does not regulate the Internet or Internet Service Providers (ISP). You may contact your state consumer protection office or if there is possible fraud involved, you may contact the Federal Trade Commission.

Having said that, http://www.fcc.gov/cgb/consumerfacts/highspeedinternet.html says the following:

If you experience a problem with your broadband service and can't resolve it with your provider, you may file a complaint with the FCC. You can file your complaint by email ([email protected]); the Internet (www.fcc.gov/cgb/complaints.html); telephone 1-888-CALL-FCC (1-888-225-5322) voice; or 1-888-TELL-FCC (1-888-835-5322) TTY; or mail:

Federal Communications Commission

Consumer & Governmental Affairs Bureau

Consumer Inquiries and Complaint Division

445 12th Street, S.W.

Washington, DC 20554.

- Valstaag

[Basil]

Before I talk about latency, let me tall you that my speeds are just fine. Even during peak hours, my speeds are almost always near their max. This post is about an even more complex issue, latency. Specifically latency of UDP traffic. I welcome all responses on this topic, so long as they are articulated intelligently.  {more omitted to save bandwidth 

I must agree with almost everything you said.

I'll add that secure connections, both browser and email have gone to the toilet.

While the FCC route you describe may have some merit it is a very protracted process.

All of us that use Satellite internet, regardless of provider, have one thing in common: we live in the boonies. Small communities. We can turn this handicap to our advantage. Word of mouth in my county is faster than WildBlue. Since I know personally by name more than half of the inhabitants in this county they ALL know how well (or badly) WildBlue is performing.

I have been with WB for well over a year and in the beginning I was recommending it to anyone that would listen. I really wanted the company to really succeed. Now I do exactly the opposite.

[Valstaag]

I must agree with almost everything you said.

I'll add that secure connections, both browser and email have gone to the toilet.

While the FCC route you describe may have some merit it is a very protracted process.

All of us that use Satellite internet, regardless of provider, have one thing in common: we live in the boonies. Small communities. We can turn this handicap to our advantage. Word of mouth in my county is faster than WildBlue. Since I know personally by name more than half of the inhabitants in this county they ALL know how well (or badly) WildBlue is performing.

I have been with WB for well over a year and in the beginning I was recommending it to anyone that would listen. I really wanted the company to really succeed. Now I do exactly the opposite.

Yes, word of mouth IS a powerful thing.  Secure connections in the browser might be helped by configuring your browser to use WB's proxy server (use address http:wpad.wildblue.com/wpad.dat in the proxy settings under 'use automatic configuration script') One thing I forgot to mention is their FAP. While their FAP thresholds are fairly generous, from what I've seen their usage monitoring has major problems. If you're going to have such a policy, you really need to back it up with accurate monitoring for both the customer and the service provider, imho.

-Valstaag