Comments on: Why ATG’s Core Based Licensing is Stupid

By: Devon

Devon — Thu, 15 Jul 2010 20:42:12 +0000

Yeah. If ATG moved to a “percentage of transactional revenue” model, you could go virtual, and do some very cool auto-scaling cloud stuff. That would be killer for SaaS/Hosted Storefront offerings.

I’m not holding my breath though.

By: Justin Commenting

Justin Commenting — Thu, 15 Jul 2010 16:01:58 +0000

Virtual machines throw a big wrench in the per-core pricing model too.

By: Hein Bouman

Hein Bouman — Sat, 15 May 2010 01:46:00 +0000

Hmmm. Maybe I have to re-read again the material. Anyhow, with SQL-server that runs on virtual environments you run in another challenge. Moving your system around to another node can suddenly change your licensing requirements. It’s done before you know. I must Microsoft is much better in their licensing structure than Oracle: http://hein-bouman.blogspot.com/2010/05/software-licenses-should-be-based-upon.html

By: Devon

Devon — Wed, 12 May 2010 14:20:00 +0000

Hein,

Oracle Standard is thankfully socket based, although you’re right Oracle Enterprise is core based, although at least they have a core multiplier, so you aren’t paying full price per core.

The problem with transaction based is there are some applications which will run a high number of transactions but they could be very light or async and don’t need a ton of HW, and/or aren’t related to making any real money, etc… It’s a tough one. ATG, being primary an eCommerce platform could charge based on on-site annual revenue, which would scale nicely and be pretty fair.

By: Hein Bouman

Hein Bouman — Wed, 12 May 2010 09:28:13 +0000

Love the rant, but sorry to say that Oracle still charges per core. At least for the app server and the database. I am just writing something for my own blog about this and did some searches to see what others already wrote. So I ended up here. I think licensing should be on a per usage basis, e.g. occurrences of use or amount of transactions.

By: Devon

Devon — Thu, 25 Mar 2010 12:34:56 +0000

Thanks! I’d love to see ATG sell based on a %, but that’s hard to enforce as many customers end up rolling their own payment integrations or heavily modifying the OOB ones. The other problem is that they’d still have to bill for support, as that can be a big cost center on a platform as complex (and dare I say it – poorly documented) as ATG is. But there’s no real reason support costs would scale with the number of CPUs/Cores a company ran on, so support could be a fixed per customer annual fee or something along those lines. We’ll see.

By: Arturas Kvederis

Arturas Kvederis — Thu, 25 Mar 2010 07:13:57 +0000

Really interesting article and discussion Devon. And I personally believe that most of the future licenses will be based on a percentage of transactional revenue as more and more SaaS ecommerce solutions and providers will pop up, the competition will get worse and eventually either they will have to change their licensing policy to stay competitive, or they will start loosing customers imho.

By: Devon

Devon — Mon, 08 Feb 2010 23:05:33 +0000

I’ve often wished ATG would offer a model based on a percentage of transactional revenue on the site. This would let the smallest mom and pop site use ATG and would scale up pretty well even to the huge players. I’m not sure how difficult it would be to implement in a way that wasn’t easily circumventable (since you need to allow folks to add custom payment processors, etc…) although realistically it’s all sort of on the honor system even now. The flip side of that is you still have to bill for support differently, as staffing to provide phone support to a client who pays you $10/year is a whole other barrel of monkeys.

I put forth the socket pricing because at least it’s inline with the other vendors you’re likely to deal with (RedHat, JBoss, Oracle).

By: Harshal Wanjari

Harshal Wanjari — Mon, 08 Feb 2010 22:59:50 +0000

Great comments Devon and Mike, pretty much summarizes a more general view I’ve seen on ATG licensing in the industry.

I favor the business/transaction volume based model over anything that ties licensing down to sockets or cores. ATG commerce implementations are fundamentally different based on the type of business that they support (cater to different assortments, conversion ratios and fulfillment models) and my experience is that at large scale – a commerce implementation is more memory and instance(jvm) bound than CPU bound – this forces one to compact multiple instances per socket/core (there are limits to how far that can go) and eventually allocate more hardware/sockets.

Most large scale commerce implementations end up under-utilizing the CPU due to this and eventually requiring more CPU/cores than what’s needed for true request processing – eventually driving up license costs. Hence tying up licensing to sockets puts a customer organization at a significant commercial disadvantage – especially the ones that have to support a large number of sessions (low conversion)

My 2 cents

By: Mike McCloskey

Mike McCloskey — Mon, 08 Feb 2010 15:50:53 +0000

This is a great article Devon and pretty much dead on in terms of the issues facing ATG right now. I think that the socket approach makes the most sense as well and it doesn’t really “change” the license model that much, although from a company perspective, they would need to change the cost of a socket.

The difficulty comes in getting your BATNA out there if you are ATG. The best negotiating approach for them is to look at volume based pricing and bundling in order to offset their cost of sale and not let software out the door for a price that will hurt them in the future.

If I look at the 16 core model (I have some cost ranges on that depending on volume based pricing), you are looking at between 800k and 1.2MM for ASE + commerce. This should allow you redundancy of 2 servers running the Nehalem processor. I see this as more of a “starter kit” of 2 “sockets” which should go for between 400-600k depending on volume discounting for this piece. This would essentially cut revenue in half though due to hyper threading, but at the end of the day it probably wouldn’t.

I think people are going to spend what they are going to spend and increasing your commerce software sale for additional processing power can always be made if you utilize the volume discounting approach.

So, if I were to go with 2 sockets at 600K or I could up that to 4 sockets applying a discount for a larger purchase at 800K. If I throw in a small bundle of ATG Recs and CA and CSC for 200k, I am still at a 1MM software sale and everyone wins.

In short, I think you are right