With the typical fanfare, GigaSpaces' XAP 8 gets released today. If you want to get into the mind numbing set of cliches that every press release pumps out in spades, then you'll be wont to check out this one. Openness and continuous scaling with zero down time are some of XAP 8's big boasting points. They've also simplified administration and they're leveraging. The press release uses the term 'leveraging' several times. so you know that these people are doing some serious work with a lever and a fulcrum.

"With XAP 8.0, distributed, scalable and highly performing applications can be designed, modified, and implemented quickly while leveraging existing knowledge and technology resources."

More to the liking might be todays blog by Nati, the CTO of GigaSpaces. Addressing the question: "What would be your scorecard for second-generation PaaS and how would it be different than an App Server scorecard?" Nati comes up with a set of criteria for evaluating a good Paas solution, with elasticity, multi-tenancy and continuous deployment as perhaps the top three topics of evaluation.

TheServerSide developer might be interested in the 'diss' Nati takes on middleware services, placing it last on the list, like it was Phil Kessel at the NHL All Star Fantasy Draft. His reasoning is flawed, yet interesting nevertheless. The flaws in his logic could be easily explained away here, but like Fermat's Last Theorem, the margin here is simply too small to contain it.

Nati's Blog: Evaluation Criteria for Evaluating Paas

GigaSpaces 8 Press Release: Leveraging all that's Leveragable

The scorecard I really don't get at all.

Multi-tenancy is an implementation detail that a customer of PaaS would naturally prefer to do without. I can't imagine a customer calling up a PaaS vendor and requesting that he would like to share hardware. The point of cloud is to put hosts, servers, vm's and eventually os processes into the background very much like our telco infrastructure to some degree. Customer's don't care about other customer's data unless its about security. Again Nati should just talk directly to the PaaS vendors and not the PaaS customers.

DevOps as a score item? Seriously I should have stopped reading at this point.

Continuous deployment and Scalability have been solved in a piecemeal fashion. If we really want to see this in the cloud we are going to need it to be inherent to the whole runtime and all components and services even libraries. That's something that is never going to be achieved in the next few years so its always going to come down to the weakest link.

Funny I see no mention of security, QoS, metering & monitoring, billing & chargeback (what's the business model) which seem pretty fundamental to the quality & value of a PaaS offering. 

(full disclosure: GridGain sometimes competes with GigaSpaces)...

Bill,

I don't see GS 8.0 release and its key points as negatively as you do. PaaS is rapidly evolving concept and we can all list all the features that would ideally have in our products. No one has a full stack yet, and no one knows what a full stack should look like (but we have our educated guesses :)

It is also important to understand the history of the products and the fact that it is extremely hard to turns products like GS, GridGain, Coherence, etc. quickly around due to their size and inherit complexity. Many things will require an incremental changes from release to release - especially something like billing, multi-tenancy, deployment/provisioning, etc.

I share one believe with Nati very strongly - 2011 will be the year that Private PaaS concept will make the strong debut - and each vendors will approach it differently.

Nikita Ivanov.

GridGain Systems.

William 

See my response below..

Multi-tenancy is an implementation detail that a customer of PaaS would naturally prefer to do without. I can't imagine a customer calling up a PaaS vendor and requesting that he would like to share hardware. The point of cloud is to put hosts, servers, vm's and eventually os processes into the background very much like our telco infrastructure to some degree. Customer's don't care about other customer's data unless its about security. Again Nati should just talk directly to the PaaS vendors and not the PaaS customers.

Were targeting those who want to build thier own PaaS - which is similar in nature to those who managed application server deployments.  If your building your own PaaS you would want to be able to offer various degreess of sharing and isolation depending on the cost/SLA for the application.

DevOps as a score item? Seriously I should have stopped reading at this point.

 

Think as someone who needs to build his own PaaS. Anything in a PaaS wolrd need to be fully automated. That include the deployment, scaling, fail-over.

Continuous deployment and Scalability have been solved in a piecemeal fashion. If we really want to see this in the cloud we are going to need it to be inherent to the whole runtime and all components and services even libraries. That's something that is never going to be achieved in the next few years so its always going to come down to the weakest link.

Funny I see no mention of security, QoS, metering & monitoring, billing & chargeback (what's the business model) which seem pretty fundamental to the quality & value of a PaaS offering. 

Obvioulsy the list of items in the scorecard is not complete. I would agree that items that you listed would have been part of the full list. The main point that i was trying to make through this excersize is to show how things would be different in a PaaS world than the way we used to think in the Application Server world.

The scorecard I really don't get at all.

Multi-tenancy is an implementation detail that a customer of PaaS would naturally prefer to do without. I can't imagine a customer calling up a PaaS vendor and requesting that he would like to share hardware. The point of cloud is to put hosts, servers, vm's and eventually os processes into the background very much like our telco infrastructure to some degree. Customer's don't care about other customer's data unless its about security. Again Nati should just talk directly to the PaaS vendors and not the PaaS customers.

DevOps as a score item? Seriously I should have stopped reading at this point.

Continuous deployment and Scalability have been solved in a piecemeal fashion. If we really want to see this in the cloud we are going to need it to be inherent to the whole runtime and all components and services even libraries. That's something that is never going to be achieved in the next few years so its always going to come down to the weakest link.

Funny I see no mention of security, QoS, metering & monitoring, billing & chargeback (what's the business model) which seem pretty fundamental to the quality & value of a PaaS offering. 

Fair enough, but I'd like to respond, because I think the points are valid but answerable.

First off, multitenancy is ... crucial to a platform-as-a-service. Nati doesn't mean that it needs to be thrust in one's face with neon lights and perhaps an alarm bell or something like that. It means that multi-tenancy needs to be available, supported, known. Being able to support multitenancy doesn't mean what you seem to presume.

DevOps should also factor in, because that's the initialization point of any service. If deployment support is difficult or difficult to control, then that enhances the cost of service development and deployment dramatically. While one hopes that deployment artifacts are always perfect, they're just not. Ignoring this facet when you consider a platform is madness. (Why do you think so many application servers focus on the user experience?)

I agree: Continuous deployment and scalability have been solved in a "piecemeal fashion." With all due respect, that's the problem. Not considering them as a whole - along with everything else - is part of the problem that a scorecard like this is meant to help solve. Piecemeal architecture, piecemeal solutions of scalability... sure, they work, but they end up creating lots of opportunity for companies like JInspired, shall we say. Not that I resent that, but there should be - and is - a better way.

Security is a good point, honestly. QoS was addressed (just not with that term); metering and monitoring were as well (in the devops bit you criticized, naturally), and billing and chargeback are concerns that I'm not really able to address.

PaaS is dead on arrival if you ask me in how it is being defined by various vendors in the cloud space that are looking for complete platform lock-in (vForce) on a scale that was never achieved by Java EE application servers or those flogging a framework that simply sugar coats layers upon layers on top of standard enterprise technologies.

We need to start thinking in terms of services and service exchanges, activities and activity streams, tasks and execution containers, mobility with context and intelligence. PaaS 1.0 is largely focused on Apps and its all about limiting choice for immediate gratification pretty much giving up all control and knowledge acquisition. OK maybe for Rails apps but enterprise apps I think not.

I understand there will be a need for particular execution points in the cloud offer proprietary container environments but these are part of a whole - a whole that is not targeting one platform or one vendor or even one cloud.

PaaS is Silverstream & NetDynamics all over again (pre-J2EE) and if people have any experience they will know that the ending is not going to be nice.

PaaS is Silverstream & NetDynamics all over again (pre-J2EE)

Dude... feels like it was in previous life...

I recently posted this comment on HighScalability on PaaS 1.0.

PaaS 1.0 is like the public telephone booths of yesteryear. Its starts out terribly expensive but all new, shiny & clean. Very quickly it becomes soiled by its multi-tenancy, lack of policing & maintenance, its limited (cramped) space & capacity, its lack of adaptability (upgrading is a bitch because its controlled by a single authority) to innovations elsewhere, and then there is all those business (card) adverts for questionable services plastered all over it. What's amazing you continue to pay the same expensive call rate while it goes through this lifecycle process

Well! if you're interested in the more technical aspects of the release - i.e., the interesting bits - check this post out: XAP 8.0 Released – Same Data, Any API and Continuous Scaling.

Quick overview: we've added JPA and a schemaless API (a document API), including some very strong nested object query and indexing capabilities, and we've upgraded our scalability features by quite a bit (faster synchronization and replication, release compatibility among future versions, and our elastic service manager is getting better all the time.)

Plus: more cool features on the way. :)

Joe my issue is with PaaS and not GigaSpaces though I think the article should really have been directed more explicitly at those thinking of building their own PaaS offering rather than those deploying to a PaaS environment. All the points Nati makes are valid when redirected so but from a customer of PaaS I am not sure it makes much of a difference if they are heading the PaaS route they are pretty much already blind folded to the realties of the cloud.

I would prefer vendors such as yourself focus more on offer specialized containers that integrate into a multi/composite service supply chain which I think is were we are headed with the cloud - exchanges, brokerages, (inter)mediators,....

I would prefer vendors such as yourself focus more on offer specialized containers that integrate into a multi/composite service supply chain which I think is were we are headed with the cloud - exchanges, brokerages, (inter)mediators,...

I don't think it's coming anytime soon. That reminds me of the same ideas of WS-based B2B exchanges...

Actually I think it has a much better chance of happening because it's not driven by a standards think tank group but by the businesses that will operate in the cloud where cost, performance and value (revenue) are are being viewed at the point of service interaction and in the resulting processing which judging by Amazon's continued innovation is all about consuming multiple metered/billable services across shared resources.

Congrats, Nati :-)

If only your software was as economically as your postings ;-)

.. hey, somebody has to make Ferraris ;-)

I should be more like you, in terms of postings naturally ;-)

Congrats, Nati :-)

Thanks Cameron

I think that it would be interesting to see the score card from the eyes of a real customer.

Just posted one on my blog here

So it leverages a number of things. But does it also shift any paradigms ?