Boost internet speed? is this possible?

E

ericmachine

n00b
Joined
Aug 14, 2012
Messages
33
Hi everyone,

Like to check. Let's assume I have 2 internet lines (1 is the main internet and 2nd is the backup, in case the 1st goes down). Both is 2MB line but different service provider.

Last week I met a supplier which claimed to invent a technology. Basically you put this box inside your network (behind router or load balancer, or can become a router too). It can boost up 30 - 50% of internet speed. Sometimes as high as 200%.

Based on one of their testing at customers' place, example

590kb/sec (before)
3.8MB/sec (after)

Another example... If you have 8MB, it can even boost up to 15MB.

At the same time, it can do bonding for multiple internet lines and making into 1 big channel.

The supplier claimed it's not hacking but rather optimizing the TCPIP. But I am curious, since the box is on my network and I got a fix bandwdith from my internet service provider, how could one make a boost like this? Furthermore this is consistent boost and leverage on ANY internet service provider.

Is there an open source equivalent to such solution?

Any help from this forum? I just want to know whether this claims are possible.
 
If this was possible, I could only imagine they'd be about as rich as the guy who cured cancer.
 
Anyone who is telling you they can "optimize" TCP/IP to push more bandwidth than a circuit can handle is absolutely full of shit.

TCP/IP are protocols that introduce overhead. IP is necessary for routing, TCP is necessary for ensuring data reliability. Both reduce your total data throughput slightly (figure about 20% total on average) because of the way they work. What this means is that when testing raw data throughput, you'd never see 100% of the bandwidth the provider is providing you. To say that tweaking TCP/IP could increase bandwidth to be greater than the total throughput capacity of the circuit is ridiculous, you might decrease the overhead by, say, using larger TCP window sizes on a low-loss link, but you can never have more throughput than you would if you were able to throw out those protocols entirely.

It's also easy to "prove" a technology like this on circuits that have a burst commitment, as the throughput test is usually short enough to place the data transfer within the burst. And this is why providers have the burst... to make you all happy with your internet connection because your bandwidth on speedtest.net is faster than what they're selling you. In fact, you're really just testing your burst speed, and if you were able to run the test for longer, you'd see your true bandwidth.

All this said, WAN accelerators ARE able to increase data throughput via realtime compression. This is hardly a new technology, but only recently has it really hit viability because the compression delay in the past has been enough to cause issues. Modern-day WAN accelerators are supposedly capable of doing compression/decompression quickly enough that it's virtually transparent. I don't have any personal experience with them however, so I'm simply repeating what I've heard.
 
To say that tweaking TCP/IP could increase bandwidth to be greater than the total throughput capacity of the circuit is ridiculous

>> That was what they claimed :)

Example, another claim

leased line 2MB (guarantee stability, probably get 1.7MB actual bandwidth, uptime/downtime guarantee) vs broadband 2MB (may not be stable, due to hops and get around the same 1.7MB, downtime normally 2 business day here)

They claim putting their box in our network example with the broadband 2MB, not only we enjoy the same quality of leased line but can boost up to 4MB line consistent.

They even claimed big giants tested their product and saying their product is not seen in any competitors and really amazed them.

I think i will get a loan unit and test myself, see whether this claims are true or not. If it is real, this is "Amazing".
 
Part of me makes me think its just a device that fools speedtest sites, or holds back data and then pushes it out in bursts to make the max data rate higher.
 
450 said:
Part of me makes me think its just a device that fools speedtest sites, or holds back data and then pushes it out in bursts to make the max data rate higher.
Click to expand...

I saw that when I was messing around with the threat detection settings in Untangle. The untangle box would download the entire speedtest file, scan it for viruses, then send it off to my PC over gigabit ethernet. Voila gigabit internet speeds!... and like 5 second latency....

There was also an old Xbox Live bug, where if you used a laptop to PC with two NICs with the "share internet connection" feature enabled, your Xbox would report back to xbox live that your LAN speed was your internet speed, thus pretty much guaranteeing you hosted every session.
 
ericmachine said:
To say that tweaking TCP/IP could increase bandwidth to be greater than the total throughput capacity of the circuit is ridiculous

>> That was what they claimed :)

Example, another claim

leased line 2MB (guarantee stability, probably get 1.7MB actual bandwidth, uptime/downtime guarantee) vs broadband 2MB (may not be stable, due to hops and get around the same 1.7MB, downtime normally 2 business day here)

They claim putting their box in our network example with the broadband 2MB, not only we enjoy the same quality of leased line but can boost up to 4MB line consistent.

They even claimed big giants tested their product and saying their product is not seen in any competitors and really amazed them.

I think i will get a loan unit and test myself, see whether this claims are true or not. If it is real, this is "Amazing".
Click to expand...

It definitely sounds like they're using a WAN accelerator. It's really nothing special, though I'm sure they want you to think it is. WAN accelerators often employ TCP window size adjustments, ack delay changes, real-time compression, and local-caching options to reduce overhead, maximize throughput, and prevent bandwidth even being used for frequently-accessed online images and files. A lot of people don't know much about them (and it sounds like their own salesmen don't really get it) but it's certainly worth trying out.

It just sounds like the people trying to sell you the product don't really understand it, and are hyping it up as something that only they know how to do. Classic sales bullshit.
 
For home use, I've found pfSense to help my internet be much faster and more reliable than a normal router. It does have firewall and caching/proxy capability. I'm running mine on an Atom. It also has a lot of packages for other purposes. I know some are using it for business.
 
Thanks everyone for replying back.

I saw that when I was messing around with the threat detection settings in Untangle. The untangle box would download the entire speedtest file, scan it for viruses, then send it off to my PC over gigabit ethernet. Voila gigabit internet speeds!... and like 5 second latency....

There was also an old Xbox Live bug, where if you used a laptop to PC with two NICs with the "share internet connection" feature enabled, your Xbox would report back to xbox live that your LAN speed was your internet speed, thus pretty much guaranteeing you hosted every session.
Click to expand...

If that's how they do it, that's cheating. Is there any way or tools I can use to see whether they are cheating or not? Other than having to press speedtest.net button every time for 1 hour and divide to get the average speed?

It definitely sounds like they're using a WAN accelerator. It's really nothing special, though I'm sure they want you to think it is. WAN accelerators often employ TCP window size adjustments, ack delay changes, real-time compression, and local-caching options to reduce overhead, maximize throughput, and prevent bandwidth even being used for frequently-accessed online images and files. A lot of people don't know much about them (and it sounds like their own salesmen don't really get it) but it's certainly worth trying out.

It just sounds like the people trying to sell you the product don't really understand it, and are hyping it up as something that only they know how to do. Classic sales bullshit.
Click to expand...

But how could a WAN accelerator like you mention can boost up the internet speed? In terms of optimize the 10MB line e.g. from the service provider makes sense to me. But owning a 10MB line but getting 20MB line speed + reliability of leased line kinda shocked me!
 
Basic computer science tells you there are 2 ways you can "improve" information transmission, and only at the cost of (storage) space or (CPU/latency) time: caching and compression.

Your example of going from 590 _kilobits_/s to 3.8 _megabytes_/s means two things: 1) The supplier is full of shit and making up test cases where they transfer 98% redundant data through an efficient compressor or 2) you or they don't know whether they actually talk about bits or bytes.

Yes, you can improve throughput using a VPN that utilizes compression to an endpoint that has a higher throughput than the connection to it, and you can reduce unnecessary data transmissions using a caching proxy.

That they try to sell it as a TCP/IP optimization, though, hints to the fact that they're full of shit or rather their poor engineering department is too detached from their clueless idiotic marketing drones.
 
Yes, you can improve throughput using a VPN that utilizes compression to an endpoint that has a higher throughput than the connection to it
Click to expand...

so you mean, if my line is 2MB and by connecting to the above VPN and when I do speed test, it will show 4MB line due to the higher throughput at the endpoint? Is that what you meant?
 
ericmachine said:
so you mean, if my line is 2MB and by connecting to the above VPN and when I do speed test, it will show 4MB line due to the higher throughput at the endpoint? Is that what you meant?
Click to expand...

Yes, if the data is 50% compressible, a 2M line would show as 4M.
 
WAN optimization? Sounds like Silverpeak or something along those lines. This is certainly nothing new.
 
You must log in or register to reply here.
Back
Top