On Mon, 2011-11-14 at 13:14 +0100, [email protected] wrote: > Hi John, > > thanks for your answer. > > Could you please provide some sources for > > "and any tweaks one can make to the IP stack or the SSH protocol. " > > ? > > Regards, > > Stefan
On high bandwidth, high latency links, one generally wants to enlarge the windows size to help compensate somewhat for the latency. I've never done a packet size analysis on X2Go but I would guess this could help with large screen transfers. Honestly, I've never taken a look at what WAN optimizations are available for SSH. If anyone knows of any, please share them. Thanks - John > > >>> "John A. Sullivan III" <[email protected]> 12.11.2011 > 03:28 >>> > > On Fri, 2011-11-11 at 14:21 +0100, [email protected] wrote: > > Hi, > > > > we are trying to use X2Go to provide a devolpment environment for > our > > employees in India. The problem is that the connection to India is > > really bad (latency ~200ms) and so working with X2Go is currently > not > > possible. > > > > Is there a way to improve the performance on the X2Go client/server > > configuration? Does anybody have experience with such a szenario? > > > > Please help. > <snip> > The first thought is the usual WAN improvements such as WAN > accelerators > and any tweaks one can make to the IP stack or the SSH protocol. Also > try to use hard wired connections on the client side rather than > wireless. We have seen a tangible difference on the same Internet > connection between the two. > > However, I am beginning to wonder if something important has changed. > Our experience with the older 3.0.1 series clients was that they were > adequate with 200ms latency and packet loss up to 4-5%. We are still > running the older server but have started using the newer client. I'm > currently using the version in the heuler repository. I've noticed > different and troubling behavior which I am guessing is a result of > the > switch to libssh. > > My Internet connection has been a little less stable than usual lately > and I've started to notice really problematic behavior. I do not know > why but, if we start experiencing moderate packet loss or latency, the > behavior is almost as if nagle is enabled. We'll type a series of key > strokes and they will all type but the last one. Two seconds later, > the > last key appears - makes editing documents a dangerous nightmare as > the > cursor is usually one character off from where one thinks it is. We > also see major screen painting delays - clicking on some action and we > know the server has responded instantly but the old screen just sits > there for two or three seconds and then suddenly changes. > > We've not had the time to track it down and I'm guessing it will be a > bear to find but something has definitely changed in the newer client > code and not for the better - John > > > _______________________________________________ X2go-User mailing list [email protected] https://lists.berlios.de/mailman/listinfo/x2go-user
