You didn't answer Michelle's question. But thanks for answering mine.

P.S. Am pretty sure I know how it works.

macster,

The key here is to think of EPC as a set of functions not a set of elements.  Once you can define a set of functions you can create a software design that can instantiate the service required of an EPC.  Then it can run on commercial x86 with all the tools that come with that.

Have a look at Optus Presss Release from last week, this is not simulation, this is in a Production Network Pilot as we speak...

Innovation comes from challenging the status quo, I am sure you would agree...and yes control & data plane.

Cheers

Hi Macster,

No, I don't think this idea appeals to them at all, to put it lightly. 

And those are good questions. I'll chase those up and see what I can find out. Tks!

Doesn't NSN's Liquid Core do this sort of virtualization?

Cisco with their ASR5k. HW with UGW/USN. E/// have very new EPC products (SSR and MKViii).

Can you really see these vendors promoting anything using "standard hardware" or whatever it's called?

It's not an easy feat to take EPC elements and have, say, an MME module or package that is essentially platform-independant. Hat off to them if they do it. But are they just developing a simulation product? Or actual elements (S-Gw, etc.) that are platform independent? Are they doing both control/signalling plane and data plane? Or just the former? 

Paul,

I guess I question the virtual bit not the standard hardware bit.  The turn up and turn down VMs on somebody elses hardware is going to be costly.  Doing it on your own hardware makes sense to me.  But the dynamic bit is okay but if you have hardware doing nothing, then why is it there?

Just because I think most folks here have never looked here is list pricing from Amazon:

https://aws.amazon.com/ec2/pricing/

You need to keep scrolling to get to bandwidth.  Which I have always found to be the overlooked bit.  For a 250 Gbyte session (which is what I get from Comcast with my cable modem) bandwidth is $30.  That is with no compute and no storage and that cost is per month.

seven

Seven,

I agree with you that the VM is of course not for free, however compared to the old world of 'boxes' from players such as Ericsson, Connectem's solution is dynamically scalable and works on commodity hardware, enabling quick TTM will relatively low risk and cost.  This is particularly relevant to service providers that want to take advantage of the burgeoning M2M space.  

Although they are slowing coming around, waiting for the likes of Ericsson to fully embrace the virtual core with only tie service providers' hands.

Paul

One thing to take a look at is cost.  If you are spinning up a bunch of new servers then you have to either have them sitting idle, doing something else that you won't be doing or buying them from someone.

As long as it doesn't get out of hand it could work, but seriously you guys need to look at the real costs of an AWS instance include bandwidth and storage.  I can tell you that our current costs are 1/3rd of AWS by running real servers in a real datacenter.  Our application is not greatly dynamic, so the numbers would be different for this.  

Does this mean its a bad idea?  No, especially the notion on running on standard hardware.  But I think this notion that spinning up VMs is free has to stop.  

seven

So, where do incumbent equipment vendors stand on virtualizing the packet core? Ericsson, NSN, Huawei, Cisco, Alcatel-Lucent... what's their position here?

From interactions with some carriers and OEMs, this is surely the first of many such efforts to virtualize the PC. Intel has been touting this for a while - it helps them if standard processors can take on something this complex (http://goo.gl/CRbxc). For the carriers, offering M2M and QoS depends on making the best use of available resources and lowering cost of operation. 

Michelle, great you point out that this is more than a fancy hardware / blade play powering software processes to match on the cloud, the game changer is the software in order to achieve the agility required.  It is refreshing to see Optus and maybe SingTel by default making a move,  I can imagine others with a similar point of view are not so far behind.  

With IBM in the mix the small start-up is supported by a very big and hungry systems player.