Last week I started backing the crowd funded GNS3 1.0 project by making a small(ish) donation, I watched the preview video and I was very impressed.
Some of the exciting new features of GNS3 are:
Grouping of devices for easier management
IOS on UNIX Support
IOU Switching Support
Ability to plug in network cables to running routers and switches (instead of having to turn them off)
Rapid Spanning Tree support
Enhanced Idle-PC calculations
Many of you have probably used GNS3 for some time now. some of you may even have weighed up the pros and cons of using IOU, so its great that GNS3 now offers the ability to leverage the best of both worlds in one interface, but how does it shape up to CML, which is clearly going to be it’s main (if not only rival)?
Costs and Limits
CML is going to be with us pretty soon (although Cisco are still keeping pretty quiet about it). The word on the street is that for the likes of you and I it will cost $100 and support a maximum of 15 devices.
GNS3 does not have such a cost or device limit (although it does require memory from the host machine), and the CCIE exam itself has far more than 15 devices (30 from what I have seen of the v4.0, and would guess it’s roughly the same for the v5.0).
With CML IOS switching won’t be released until later in the year, possibly as an upgrade, maybe as an add-on, unknown if there is a price association.
GNS3 Switching will be handled through connectivity to an IOU VM.
There are expected to be other issues with CML, in that as each device runs as a separate VM it is a “huge resource hog” – and this is coming from someone who worked on the beta from a thread over at the Cisco learning network here.
Both CML and GNS3 1.0 will hook into VMs for different requirements, out of the box GNS3 works exceedingly well, granted switching has been a bug-bear for most. GNS3 can already hook into VirtualBox for VMs and can run IOS XR vms, and any manner of VirtualBox guest. CMLs VM connectivity will be decided by Cisco.
CML offers a great way of overlaying data – from my earlier post on VIRL (as it was known then) you can overlay information such as BGP connections, which will be great for presentations, especially as the engine can be separated from the GUI (called VM Maestro), the GUI can run on a laptop and the engine could for instance run on a server back in the office. Both ca, however, live happily on the one laptop though (if suitably specced). This separation of roles is very good, especially if this is in a corporate environment.
GNS3 can connect to read equipment whereas CML seems to be more self contained, which also keeps it within the node limit imposed by your license, but also means that one of the great features of GNS3, that overcame the switching issue, will go back to relying on emulation of ASICs to provide switching.
Who will win?
It’s really early days yet, but for the one-off consumer looking to study for an exam I think it will be GNS3 that has the greater install base. CML will get a lot of attention, thats pretty obvious, but its device number limitation for the like of you and I that will be a real bug bear. GNS3 is free and will remain free, it has a great following and is only limited by the imagination. Yes I will will buy CML, just like I bought about $2000 worth of routers and switches before finding out that they are going to be no good for the CCIE v5, but I think I will probably use GNS3 more.