LVS
lvs-users
Google
 
Web LinuxVirtualServer.org

Re: a probelm about VS-TUN

To: Horms <horms@xxxxxxxxxxxx>
Subject: Re: a probelm about VS-TUN
Cc: lvs-users@xxxxxxxxxxxxxxxxxxxxxx
From: Wensong Zhang <wensong@xxxxxxxxxxxx>
Date: Sun, 26 Mar 2000 20:51:12 +0800 (CST)


On Thu, 23 Mar 2000, Horms wrote:

> On Fri, Mar 24, 2000 at 10:19:33AM +0800, Wensong Zhang wrote:
> > 
> > 
> > 
> > On Thu, 23 Mar 2000, He Weitao wrote:
> > 
> > >     I wander two questions:
> > > 1. how many realservers can be extended in the VS-TUN method?
> > > 
> > 
> > It depends on your network applications and your real server throughput.
> > If incoming requests carry less data (such as http service), you can have
> > a large amount of general servers (50 or 100).
> 
> I would have thought that the limit was packets/s rather than bandwidth,
> in which case the ratio if incoming to outgoing traffic wouldn't
> be som important. Does this make sense or is bandwidth the major
> factor?
> 

For the limit of the load balancer, it is good to use bandwidth to
measure. However, for the throughput of the whole cluster, it depends on
application services. For example, the bandwidth of a load balancer is X
Mbits/s, the throughput of VS/NAT based services might be around X
too. For VS/TUN or VS/DR, you probably need the ratio of outgoing to
incoming traffic (such as R) to caculate the throughput of the whole
cluster-based services, which might be X*(R+1).

However, with different hardware configuration, the bandwidth limit of a
load balancer may vary too.

Thanks,

Wensong 



<Prev in Thread] Current Thread [Next in Thread>