<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 TRANSITIONAL//EN">
<HTML>
<HEAD>
<META HTTP-EQUIV="Content-Type" CONTENT="text/html; CHARSET=UTF-8">
<META NAME="GENERATOR" CONTENT="GtkHTML/4.4.4">
</HEAD>
<BODY>
On Wed, 2013-06-05 at 20:05 +0300, Mihaela Alexandra Martinas wrote:
<BR>
<BLOCKQUOTE TYPE=CITE>
I started analyzing how the multi-engine can be designed and a few questions popped up:
</BLOCKQUOTE>
<BLOCKQUOTE TYPE=CITE>
1. The user should have access to the cli of the switches that the multi-engine attempts to hide?(I know that the purpose for creating the multi-engine was that the user to access in a transparent manner the switches. Following this logic to access the clis shouldn't be necessary).<BR>
</BLOCKQUOTE>
I would say "not necessarily", but it would be a "nice-to-have" feature, though. In other words, if having this feature complicates the design and bloats the code, then forget it.<BR>
<BR>
<BLOCKQUOTE TYPE=CITE>
2. Since the switches are on the same machine is it ok to communicate through sockets with the hidden switches?
</BLOCKQUOTE>
No, it doesn't make sense to use sockets if the switches are on the same machine. But on the other hand, who says the switches are on the same machine? Maybe you have a cluster of VM hosts, each of them having a bridge+8021q lisa switch. And you want to deal with the whole cluster as if it shared a single big switch.<BR>
<BR>
If the design is based on sockets, it can easily scale to a distributed system but can still be used on a single-machine system. By the way, sockets on the loopback interface are really fast ;)<BR>
<BR>
Radu<BR>
<BR>
<BR>
</BODY>
</HTML>