View Single Post
Old 02-15-2012, 01:26 PM  
Brad Mitchell
Confirmed User
 
Brad Mitchell's Avatar
 
Industry Role:
Join Date: Nov 2001
Location: Southfield, MI
Posts: 9,812
Quote:
Originally Posted by HomerSimpson View Post
their demo speeds are ridiculously slow..
Honestly, I've never heard that before.. but you were right - thank you! That test server you downloaded from was in Phoenix and we just changed the links to MojoHost in Miami for downloads. I would politely ask you to re-test. No fault of that host, I just discovered in the last hour its our mistake for linking from that server. A developer was, well, developing and its config was whacked yesterday.

FastTCP is a fully compatible replacement for the TCP/ip stack on servers that use the following operating systems:

SLES10 SP4
CentOS/RedHat 5.7
CentOS/Redhat 5.4
Ubuntu 8.04
Ubuntu 10.04
FreeBSD 7.2

Specifically, what to compare using the download test at www.fasttcp.com/demonstration is not the exact throughput you receive, rather, the difference between a download accelerated that uses FastTCP versus the unaccelerated download that is regular TCP/ip.

In the coming days (when we update our website html), you will be able to do downloads from more than MojoHost in Miami; Quadranet (Dallas), Swiftco.net (Seattle), SecuredServers/Cavecreek (Phoenix), Webair (New York), LeaseWeb (Amsterdam), AdultEUHost (Amsterdam) and XRNetworks (Ashburn). While results vary from any point to any other point on the internet, FastTCP is always faster than regular TCP. In most instances this eliminates the download performance gap that people aim to accomplish by using a CDN.


Here is a better explanation from Chris O'Connell:

FastTCP Technical Overview

Christopher O?Connell
VP of Technology
[email protected]

At it?s heart, FastTCP 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 Reno1, 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 communicate2. 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 Fast.

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.

In contrast, FastTCP 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, FastTCP optimizes the connection
by slowing just enough that a catastrophic packet loss never occurs. FastTCP therefore
sends data at the maximum possible speed for the maximum amount of time.

Since FastTCP 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 FastTCP 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.

-------------
We are happy to field any more questions on the product, wherever they come in from. Nearly every host familiar within "this" space will be offering this to you in the next 30-90 days. FastTCP is a separate company than MojoHost. If you don't want to talk to us at MojoHost or FastTCP directly, I encourage you to ask your existing host. Not all host company employees will be familiar but a majority of owner/operators have heard or are already in discussions to re-distribute.

Cheers,

Brad Mitchell
__________________
President at MojoHost | brad at mojohost dot com | Skype MojoHostBrad
71 industry awards for hosting and professional excellence since 1999
Brad Mitchell is offline   Share thread on Digg Share thread on Twitter Share thread on Reddit Share thread on Facebook Reply With Quote