Thread: sendfaster.com [speeding up sites]--> bullshit or good thing?

It's not an outright scam in that there is some theory behind it. But I would be very hesitant to pay that amount unless I were making many thousands per server in profit where even a 1% improvement would pay for it.

Wikipedia article: http://en.wikipedia.org/wiki/FAST_TCP

It looks like it was developed about a decade ago. Since that time I believe the standard TCP stacks used have improved over the years to make the difference much less.

http://en.wikipedia.org/wiki/TCP_con...ance_algorithm (I think most Linux kernels are using CUBIC now?)

Additionally the largest webhosting forum I know of, webhostingtalk, only has less than half a dozen posts about Fast TCP. Take from that what you will.

The tricky thing is that with some hosts of less repute they might try to shakedown a customer after they cancel such a service by purposely degrading performance. Then when you call up to complain or ask what's wrong they might say "We find nothing wrong. Maybe it's due to the loss of Fast TCP?" Mojohost has a good reputation. I very much doubt they'd do this but I thought I'd mention it anyway.

I'm not an expert though so see what others have to say too. I last studied TCP/IP protocols over 12 years ago at university and am more than a bit rusty!

Originally Posted by matyko

Thank You Very Much for the Lots of valuable info. I am canceling this service...

Peace,

Matyko

Matyko,

I figured when you cancelled it was because you couldn't afford it, not because you were mis-informed. AllNiches made an attempt to educate you which is very nice and effortful. However, as he says in his post he's not an expert. Clearly, not familiar with the technology other than a Google of it and hasn't actually done any testing himself. I'm a little disappointed that we didn't talk about this directly. I need to help you have a better understanding of its use and benefits.

Originally Posted by sandman

looks interesting but im not sure its worth $99 a month

It depends if you're a hobbyist online or here with a full time business and income. The benefits to using SendFaster are just about the same from a primary server than if one was to use a content delivery network. With complete compatibility and nothing required for the end surfers to receive the benefit, for most it's a no brainer.

Originally Posted by allniche

It's not an outright scam in that there is some theory behind it. But I would be very hesitant to pay that amount unless I were making many thousands per server in profit where even a 1% improvement would pay for it.

Wikipedia article: http://en.wikipedia.org/wiki/FAST_TCP

It looks like it was developed about a decade ago. Since that time I believe the standard TCP stacks used have improved over the years to make the difference much less.

http://en.wikipedia.org/wiki/TCP_con...ance_algorithm (I think most Linux kernels are using CUBIC now?)

Additionally the largest webhosting forum I know of, webhostingtalk, only has less than half a dozen posts about Fast TCP. Take from that what you will.

The tricky thing is that with some hosts of less repute they might try to shakedown a customer after they cancel such a service by purposely degrading performance. Then when you call up to complain or ask what's wrong they might say "We find nothing wrong. Maybe it's due to the loss of Fast TCP?" Mojohost has a good reputation. I very much doubt they'd do this but I thought I'd mention it anyway.

I'm not an expert though so see what others have to say too. I last studied TCP/IP protocols over 12 years ago at university and am more than a bit rusty!

None of the other stacks come close to this implementation of TCP in performance, lest of all anything in a standard implementation. FastTCP is patented and has always been the most stable algoryhthm with highest performance where even the best of other options are still truly more erratic. This isn't a difference of 1% improvement. The average throughput increase is 15% and the common result is 50-100% better throughput. Even though this isn't something we have announced or made generally available, there are currently dozens of active servers at MojoHost and elsewhere using our distribution. I didn't realize this was on the boards or I would have posted sooner. Satisfied customers in our pre-release include Backroom Casting Couch, PaperStreetCash, CraziesCash, Rabbits Reviews, Wasteland, Clips4Sale, Pornstar Dollars, MyGayCash, Homegrown Video and many others.

There are few posts on WebHostingtalk because it hasn't been widely successful in its commercialization. We are excited to be bringing this product to market at an affordable per server monthly license entitlement. Matytko was recently part of an early implementation that some MojoHost customers participated in.

Finally, there are no other hosts offering this - YET. Those that will in the near future will be in agreement with my new company SendFaster, just as MojoHost is. In the coming weeks this will be offered in our commercial edition to all MojoHost customers and be for resold by dozens of competitive hosting companies, nearly all of the ones this audience is familiar with and many more. While its web site is due for publishing of final revisions, I still welcome everyone to do test downloads themselves at http://www.sendfaster.com/demonstration. The 100m and 800m test files aren't linked at the moment but will be re-added early in the week. Links to geographically located servers that we have been using all year will be included in this next revision. You'll be able to test performance "accelerated vs non-accelerated" on entry level servers we pay for at SecuredServers in Phoenix, Quadranet in LA, Webair in NY, Leaseweb in Amsterdam, XR Networks in Ashburn and of course MojoHost in Miami. The only difference between each of the two links is that we have designated one port on the same server to be non-accelerated (443) and use regular TCP/IP.

I'm happy to discuss the technology further by phone call and of course email to brad@sendfaster.com or brad@mojohost.com. I'm very exciting to be returning in the next few weeks to this board for its more proper release announcement which will include many other host companies besides MojoHost.

Enjoy your weekend!

Brad

Originally Posted by oil

i did some download test from that site with the 14 mb file
http://www.sendfaster.com/demonstration

my results are they both start out slows, and turn slow to the end,
just that the sendfaster file reached a higher peak download, and the slow down of the end was higher

but i could not realize any significant time saver on those, so my conclusion is its good if you give out large files, so proll for a paysite serving large video files faster it would helpbut besides that i dont see a benefit it this, besides that fact that i doubt the service is worth 1200 USD per year

i have a tendency to say youre way better off with CDN then this one

I'm not sure where you are physically located but there is typically a more notable impact for connections with a large latency (distance) or any amount of packet loss (re-transmissions). The effect that latency has on transmission throughput isn't as profound for short distance transmissions, typically those less than 500-1000 miles. The effect that any amount of packet loss has on throughput is substantial. The TCP/IP protocol that everyone uses doesn't have intelligent recovery. What does this mean. Well, when the protocol was developed in 1974 packet streams were infinitely smaller, flows of 100 bytes. Therefore, a packet having to re-transmit was in fact a big deal. The way the TCP/IP stack handles the loss of a packet and re-transmitting is essentially to halve the transmission output window sizing to that end user. What this means, especially in today's world, is that connections to servers are rarely optimal or reaching their potential, further, that the TCP/IP protocol does not itself have an intelligent recovery in its algorhythm. Remember, the judge of how this can benefit your business isn't what your single d/l test is. It's the cross-section of results that all connections have to your server, further, the ability to so dramatically improve results for your outlier audience - the ten to twenty percent of those most challenged by the quality of their ISP or distance.

It is certainly harder to spot the overall difference on a small file unless you're using something programmatic to measure the exact times and if you're using a single test versus repeating, well, then you definitely don't have satistically valuable data. Hit these two files, they reside on the same server and because they're 200m it should be clearer to spot:

This link is to a 200mb SendFaster TCP download:
http://sendfaster.com/static/big_buck_bunny-480p.avi

This link is to a 200mb regular TCP download (same server, port 443 is configured to use regular TCP/IP):
http://sendfaster.com:443/static/big...bunny-480p.avi

My opinion and experience varies greatly from everyone's.. obviously. I've been working with this technology for three years and have substantial experience in its implementation. We have clients who have done their own head to head testing of primary hosted servers with SendFaster versus multiple commercial CDNs on the market and subsequently chosen to drop their 3rd party CDN services. We are fortunate enough to have the "unlikely" and "underdog" solution for mass distribution. Why one would favor a more complex and expensive solution of CDN over an elegant, low cost and compatible solution for their primary server is beyond me. Two clients of mine come to mind specifically who dropped back absolutely huge savings to their budget by replacing out CDN and serving from origin. One of them went from spending $10-15k monthly to the incremental difference of buying another 1000 megabit from us and renting two more servers for 1/4 the cost. They knew all of their metrics and proved that downloads from Miami, Florida to Canada, England, Germany and Georgia were just as fast as from the CDN's caches in each georgraphy. What's to hate about that?

Originally Posted by toker

Sounds similar to technology used by KillerNic that so many sheeple gamers bought into not realizing the gains are so marginal it was laughable. I can think of much better ways to invest and actually get some worthwhile return rather then trying to yield more electrons per atom. No doubt hosts will probably begin packet shaping and throttling customers to get them to buy into this stuff if they can realize bigger profits.

Hi Toker,

I'm happy for you to challenge us. If you focus on conspiracy theories about hosts doing packet shaping, it may be your loss for missing the boat. While I know for a fact that many host networks are not properly managed at a network level and that some do traffic shaping, this is NOT the case at my company and never has been. Further, that if I ever do find an instance of a reseller shaping traffic testing results, they would be in breach of our distribution agreement. I've established service at a handful of hosts already to lend credibility to the testing process. We pay the same retail everyone else does for those servers and all at the same time, we have no access to their network, haven't asked for any favors and they have no root level access to our unmanaged rentals. Anyone can claim conspiracy because there is no doubt that the technology "sounds to good to be true." In fact, overcoming this objection is understandably the main hurdle in the sales process. Easy to resolve, though - test it out and see for yourself.

If anyone thinks that the dozens of hosts who will be offering this shortly to their customers haven't themselves had their CTOs, owners, network admins and server administrators beat the sh!t out of this technology already in an effort to disprove any of these claims... then I'm being given FAR more credit than I deserve. I can assure each and every of you that while the addition of this to your server is not for everyone, it is in fact a very special product as further evidenced by the consensus of MojoHost's competitors. One would have to be wearing tin foil if they thought for one moment dozens of companies I have competed against for the last 10 years would enter into software distribution agreements with my new entity, SendFaster, without SUBSTANTIAL merit. We have an awesome product to share with the rest of the world and I couldn't be more thrilled.

Our clients know their metrics and the benefits to implementation aren't at all marginal, they're substantial. In measures of improvement, you will see several things happen after implementation. For sites rich in content, you'll have an increased time on site and page views. As has always been the case, faster is better and everything we can collectively do to improve upon the user experience lends itself to them being happier and staying longer. We also quantify a decrease in bounce rates, the ratio of people that only look at a single page of the site and then leave. Overall averaged page load times will decrease as well. If you're running Google Analytics or watching your Alexa these are all things that you will find with a right enough statistical sampling (IE, run it for a month). If you track member cancellation reasons, you might see as much as an 80%+ decrease in cancellations due to site performance as one of our customers did.

It's perhaps more entertaining for some to sit and post on the board saying "I can't hear you, I can't hear you"

I get it, I myself am a skeptic. Hell, my now business partner in the venture is wayyyyyyyy smarter than I am and his first response just like my business partner in MojoHost was the same, absolute but unqualified rejection. At the end of the day, we can all argue until we're blue in the face about results after reading some white papers and internet commentary. It's my job to be patient and informative. Nobody can argue with results. 25-50% more server output upon reboot for all of the reasons which create this (> time on site, > throughput, etc) is amazing.

But you have to ask yourself, what's the credibility of the person (me) delivering the message? I've been deeply committed to and giving back so much to our industry since 1999. We serve 700 adult companies representing far more than 50k web sites on around 1500 servers. To suggest that I'd put everything on the line at this point in my career to sell some snake oil is illogical. I've stumbled upon an amazing technology and worked very hard to obtain distribution rights and create the proper methodology to successfully bring it to market en masse. If any of you think for a New York second that every single license entitlement we have out there already was blindly led into their purchase and doesn't see an ROI for making it, I would say a humility check is in order and that dangerously high self opinions are usually quite costly in business.

Originally Posted by sandman

upgrading a server for $99 a month adding some ssd drives would help performace for most people more then a better tcp stack

that being said does this tech support freebsd 9 or is this a linux only thing ?

if this works on freebsd 9 i would love to test it out see how it works

Nothing you can do to a server's hardware will ever help performance like changing out the TCP stack. The best solutions always include the best hardware, of course, the quality of the hardware is relative to performance. It is not, however, relative to how the TCP stack is managed and how much throughput is received by end users. Again, disregard your preconceived notions and try it yourself. I'm happy to let you prove it for yourself. While we don't have our release for FreeBSD 9 yet, it is forthcoming and I will be sure to let you know when it is available. If you would like to try this on CentOS v5.X today, I'm happy to provide licensing and installation for testing purposes.


All the best,

Brad

Originally Posted by AmateurFlix

The above is from a salesman who in 2004 owned a hosting business and didn't understand what a 404 error was.

No idea (and don't really care) whether sendfaster is of merit or not, or whether he's now become something other than an investor/salesman.

Way to hold a grudge for 8 years! I hope you and your business are doing very well.

Brad

Originally Posted by toker

http://www.igvita.com/2012/01/18/bui...-realtime-web/

That's a good read and very interesting. It's related, however, not relevant.

Brad

Originally Posted by toker

I bet every solution that happens to be open source aren't relevant to your profit margin either right?

For a webmaster board themed "Knowledge is Power" and so diametrically opposed to the mannerisms and content value over at GFY, I'm shocked by some of the responses here. Maybe re-read my posts and make an effort to understand what is being communicated. I was very precise. That link WAS a good read and technically interesting, I was paying you a compliment. I then said it's related, not relevant or a replacement. It works well WITH technology. Here is an article that discusses:

http://www.theregister.co.uk/2011/08..._modpagespeed/

SendFaster's (FastTCP) purpose is to "overcome issues such as packet loss, latency, and congestion."

Originally Posted by oil

i tested downloads from the possible worst spot on this planet which is chiang mai, thailand,
however the difference on the 200 mb files is indeed huge,
my isp offers 2 MB/s download speed while on the sendfaster file i did had a download of 2.8 MB/s i never had more then 1.8 MB /s on the normal file,
needless to say the sendfaster was downloaded much quicker thou, however 100 dollars / month per server is quite a high price tag if you ask me,
the reason that method is not soo much known is prolly the insane pricing to it, as i understand things its just a software mod to the server setup,
so the reason why this method is not widespread is simply cause your main interests are to use it as a cash cow,

Thank you for testing, Oil! Please clarify for me - you experienced 56% faster throughput on your download than you have EVER seen before on your internet connection or 56% faster than the "un-accelerated" TCP/IP download link? Either way, both are fantastic results that I would reasonably expect based on your physical location.

You are certainly entitled to your opinion with regard to value proposition. A single server can push 1..10.100..500..1000..2000+ megabit. If a business can't justify $99 monthly per server for the benefit of improving throughput by 15-500%, I suppose its not our target audience. If you currently use or have considered to use any content delivery network, cloud or geographically disbursed hosting there is no question that including SendFaster as part of the strategy will yield tremendous gains in performance and savings.

This is a patented commercial product. Everyone prefers their intellectual property to be paid for, this is no different. Our price and distribution method is new. I understand this technology is unfamiliar to this audience and will better explain the history of its commercialization better. This was previously available only in enterprise markets as hardware routing devices. More recently, as a software distributable for several thousand per seat up front with high minimum quantity purchase requirements. Its benefits are proven and in use by many of the largest online companies in the world. My company SendFaster brings the kernal module to market with affordable monthly licensing.

I think our endeavor is VERY exciting for the open market. Read on please before getting riled up... again. We reduce the complexity of installation and the financial barrier to entry which previously excluded small and mid-markets. At the same time, we package new license distribution technology in addition to providing new robust end-user and channel (your web host) support options. I will leave you with the following value comparison: Wowza Media Licenses, many use this on their server. Anyone can purchase direct licenses for $995 outright or $55 monthly or through a channel (host) reseller for the range of $50-$60 monthly on average. I have more than 100 customers at MojoHost who choose monthly licensing with Wowza instead of the purchase and an 18 month ROI. We (SendFaster) resell an amazing product whose direct price is around $3000 each in quantity with no monthly pay options and requiring annual support dues. I think that our value proposition is heroic at $99 monthly for those who benefit from the implementation and use of this technology.. but that's just my opinion!

Sincerely,

Brad Mitchell

Toker,

I was clear already that we are pre commercial release. We will be on the market in July. Therefore, no references on the MojoHost web site, no SendFaster advertising, no articles or quotes elsewhere - intentionally. To share something, I did name a hand full of our existing customers as reference. The underlying patent is "FastTCP" and there is plenty of searchable information on this, including large enterprises who have implemented. My established customer started this thread and prior to my arrival, it contained mis-information. Our distribution will of course be marketed to all industries. I'm not here marketing. I am here defensively and out of respect for our client and the community which I massively serve in my other capacity as MojoHost. Enough with the negativity already. For the educational benefit of this boards readership, I suggest fully reading responses before replying.

Oil,

You raise a good academic question that I'll run up the ladder just to be sure.. However, I am certain that one would NOT expect to see any difference in ping or round trip times. Those are representative of the physical distance between two locations in milliseconds. Not being a packet teleporter, we can't change that! LOL What SendFaster's protocol does do so effectively is manage through a high (physical distance) latency intelligently. Its results are effectively always higher throughput versus TCP/IP by result of it's packet send and receive scheduling and management of the connection window sizing. Here is a transmission for illustrative purposes:

howitworks4_final.jpg

The graph illustrates the use of the SendFaster TCP protocol versus a standard TCP Reno configuration and explains why sendfaster works. SendFaster sends a fuller stream much more rapidly and the packet loss recovery provides an exponentially better end user experience in terms of page load times, jitter reduction, and overall quality of experience.

At it’s heart, SendFaster TCP is simply a more efficient and intelligent implementation of the TCP stack. In normal conversation, when people discuss TCP they refer to a commonly used implementation, TCP Reno, or one of it’s variants. Despite the almost universal adoption of Reno based variants, TCP exists as an abstract specification for the "language" with which systems communicate. Because TCP is a data interchange specification, it is possible for us to implement a superior algorithm that nonetheless meets the technical requirements of TCP. Our version is called SendFaster.

The TCP specification, in general, has one missing piece — intermediate routers cannot explicitly tell the origin that data has been delayed due to network congestion. Instead, routers delay data until their queue fills up, at which point packets are dropped. Most current TCP implementations track only whether a packet successfully arrived — the single binary digit 0 or 1. As a result, current TCP implementations send data at maximum speed until data is lost, at which point they must re-send the lost data, and radically slow their sending rate until the connection recovers. In practice, however, existing TCP implementations rarely run at full speed and spend most of the time in some form of speed limited state.

By contrast, SendFaster TCP keeps track of the time that it takes packets to arrive and computes a delay factor. As intermediate queues fill, the delay factor increases. By using the many-bits of information contained in the delay calculation, SendFaster optimizes the connection by slowing just enough that a catastrophic packet loss never occurs. SendFaster TCP therefore sends data at the maximum possible speed for the maximum amount of time.

Since SendFaster TCP uses only information already available in the TCP specification, it is 100% compatible with existing Internet infrastructure, and only needs installation on the sending end — intermediate hardware and software need not be modified. Furthermore, as SendFaster TCP does not cause the connection to cycle between too fast and too slow, it actually helps intermediate routers perform more efficiently, by not overwhelming them until the connection collapses and then slowing the connection.

Sincerely,

Brad Mitchell