Tag Archives: PaaS

Elasticity

 or   or 

(Hooke’s Law for expressing elasticity of an object in various degrees of complexity)

The equations above get pretty complex pretty quickly! And that’s when we deal with equations that have been known about for hundreds of years. When we start using elasticity to describe cloud computing, it gets even worse.

The topic was brought up the other day when I was looking at purchasing some space on the SAP HANA Cloud to run an application that we’re developing in-house. I was checking the price for this.

http://scn.sap.com/thread/3350483

I got quite confused.

Then the conversation moved to twitter and we started discussing not just the price of going to the cloud but also how it should be priced. And then even onto how it could be made multi-tenant (which is a bit beyond the scope of this post, but it was interesting nevertheless.

I think the conversation is worth preserving so I’ve made a copy of it with a little help from Aaron’s Twitter Viewer and a lot of cutting and pasting so I could do without the CSS (if anyone knows how to add custom CSS to a single WordPress post, I’d be interested.)

Have a read, it’s not a bad collection of thoughts, and interjections (by the one and only Dennis H) and I’ll recap on my thoughts at the end:

wombling
Chris Paine
Feeling slightly confused by SAPStore pricing for #saphanacloud if you understand it pls help me scn.sap.com/thread/3350483

2 days ago
1 retweets
#

rhirsch
Dick Hirsch
@wombling compare price to other #saphanacloud packages in #sapstore – all have similar structure

2 days ago
#

wombling
Chris Paine
@rhirsch I understand the free ones but still confused what calculation is for rest, why show pm price when only pa purchase possible?

2 days ago
#

rhirsch
Dick Hirsch
@wombling a good question for #sapstore and #saphanacloud team – another reason to always read the small print

2 days ago
#

vijayasankarv
Vijay Vijayasankar
@rhirsch @wombling it will get rationalized soon . @aiazkazi has plans for it

2 days ago
#

rhirsch
Dick Hirsch
“@vijayasankarv: @rhirsch @wombling it will get rationalized soon . @aiazkazi has plans for it” >> hope you guys are working on cloning him

2 days ago
#

vijayasankarv
Vijay Vijayasankar
@rhirsch @wombling hehehe @aiazkazi is one of a kind – but he has a team behind him too to help with scale 🙂

2 days ago
#

vlvl
Yariv Zur
@vijayasankarv @rhirsch @wombling @aiazkazi Pricing is presented as PM because this is how it was defined in official price list (cont)

2 days ago
#

vlvl
Yariv Zur
@vijayasankarv @rhirsch @wombling @aiazkazi (cont) however min. Contract length for all cloud subscriptions is 1 yr. hence the mess.

2 days ago
#

wombling
Chris Paine
@vlvl @vijayasankarv @rhirsch @aiazkazi certainly not the clearest situation. But then again probably simple than onPrem pricing

2 days ago
#

esjewett
Ethan Jewett
@wombling @vlvl @vijayasankarv @rhirsch @aiazkazi Minimum 1-year subscriptions are not very cloudy. Are add-on resources more flexible?

2 days ago
#

wombling
Chris Paine
@esjewett @vlvl @vijayasankarv @rhirsch @aiazkazi had one potential customer only needed 3-4 months every yr. They didn’t sign up 🙁

2 days ago
#

wombling
Chris Paine
@esjewett @vlvl @vijayasankarv @rhirsch @aiazkazi cloud ideal for flexibility, but not so much in this case

2 days ago
#

esjewett
Ethan Jewett
@wombling @vlvl @vijayasankarv @rhirsch @aiazkazi Really, I’d argue that it’s not even cloud if it requires a 1-year commitment. Hosting.

2 days ago
1 retweets
#

wombling
Chris Paine
@esjewett @vlvl @vijayasankarv @rhirsch @aiazkazi different times for use-cases SuccessFactors 3yr contract. But wld like more flexible PaaS

2 days ago
#

esjewett
Ethan Jewett
@wombling @vlvl @vijayasankarv @rhirsch @aiazkazi Indeed, but for IaaS and PaaS I’d argue “cloud” involves elasticity. The NIST agrees 🙂

2 days ago
#

vijayasankarv
Vijay Vijayasankar
@esjewett @wombling @vlvl @rhirsch @aiazkazi other than on iaaS (rhymes with aiaz) , I doubt perfect elasticity will happen for any vendor

2 days ago
#

esjewett
Ethan Jewett
@vijayasankarv @wombling @vlvl @rhirsch @aiazkazi Every PaaS I’m aware of provides it. Elastic Beanstalk, Heroku, CloudBees come to mind.

2 days ago
#

vijayasankarv
Vijay Vijayasankar
@esjewett @wombling @vlvl @rhirsch @aiazkazi maybe PaaS will get there too at some point , but seriously doubt SaaS will

2 days ago
#

esjewett
Ethan Jewett
@vijayasankarv @wombling @vlvl @rhirsch @aiazkazi Agree though that it’s not as key for applications. But we’re talking about PaaS, I think?

2 days ago
#

vijayasankarv
Vijay Vijayasankar
@esjewett @wombling @vlvl @rhirsch @aiazkazi PaaS ideally should have no lock in – just a question of how much scale justifies it

2 days ago
#

wombling
Chris Paine
@vijayasankarv @esjewett @vlvl @rhirsch @aiazkazi GApps, Azure, CloudBees PaaS are monthly, why not #saphanacloud?

2 days ago
#

vijayasankarv
Vijay Vijayasankar
@wombling @esjewett @vlvl @rhirsch @aiazkazi elasticity is definitely something on top of the agenda . Question – is monthly good enough?

2 days ago
#

wombling
Chris Paine
@vijayasankarv @esjewett @vlvl @rhirsch @aiazkazi Simple fixed CPU/data monthly makes sense, more elastic, then GApps style usage payment

2 days ago
#

dahowlett
Dennis Howlett
@wombling Isn’t the fundamental qu something like: ‘Why does #SAP find it necessary to invent new ways to confuse?

2 days ago
#

esjewett
Ethan Jewett
@vijayasankarv @wombling @vlvl @rhirsch @aiazkazi Daily or hourly would be better, but one step at a time.

2 days ago
#

esjewett
Ethan Jewett
@vijayasankarv @wombling @vlvl @rhirsch @aiazkazi Each decrease in granularity enables different scenarios. E.g. daily helps w/ month-end.

2 days ago
#

vijayasankarv
Vijay Vijayasankar
@esjewett @wombling @vlvl @rhirsch @aiazkazi what is your absolute best case granularity ? And is monthly a good enough alternative ?

2 days ago
#

wombling
Chris Paine
@vijayasankarv @esjewett best case is on demand pay as you use eg cloud.google.com/pricing/ low base price (monthly) then as needed – elastic

2 days ago
#

esjewett
Ethan Jewett
@vijayasankarv @wombling @vlvl @rhirsch @aiazkazi Hourly is kind of industry standard, though monthly is fairly common for PaaS.

2 days ago
#

esjewett
Ethan Jewett
@vijayasankarv @wombling @vlvl @rhirsch @aiazkazi With PaaS, the case could be made for value in even more granular metering than hr.

2 days ago
#

esjewett
Ethan Jewett
@vijayasankarv @wombling @vlvl @rhirsch @aiazkazi But I’d say if you can get it to hourly that’d be great.

2 days ago
#

vlvl
Yariv Zur
@esjewett @vijayasankarv @wombling @rhirsch @aiazkazi one more point – full elasticity is good for techies but hard on the CFO. (Cont)

2 days ago
#

vlvl
Yariv Zur
@esjewett @vijayasankarv @wombling @rhirsch @aiazkazi they need the ability to forecast expenses. So for DEV we have full elasticity (free!)

2 days ago
#

vlvl
Yariv Zur
@esjewett @vijayasankarv @wombling @rhirsch @aiazkazi for PROD you pay in advance for 1 yr, becoming acceptable for CFO #hanacloudportal

2 days ago
#

esjewett
Ethan Jewett
@vlvl @vijayasankarv @wombling @rhirsch @aiazkazi Good point, and I think makes sense for apps but not for the PaaS itself.

2 days ago
#

vlvl
Yariv Zur
@esjewett @vijayasankarv @wombling @rhirsch @aiazkazi PaaS is for running apps. “How much is the new supplier portal gonna cost me?”

2 days ago
#

esjewett
Ethan Jewett
@vlvl @vijayasankarv @wombling @rhirsch @aiazkazi But usually PaaS is for dev to run apps. SAP’s take seems to be that cust manages PaaS.

2 days ago
#

vlvl
Yariv Zur
@esjewett @vijayasankarv @wombling @rhirsch @aiazkazi IT manages PaaS, but the app is for the cust. Not for the DEV guy 🙂

2 days ago
#

vijayasankarv
Vijay Vijayasankar
@esjewett @vlvl @wombling @rhirsch @aiazkazi there are 2 broad uses – 1. custom development by a customer for their use and ..

2 days ago
#

vijayasankarv
Vijay Vijayasankar
@esjewett @vlvl @wombling @rhirsch @aiazkazi ..and 2. An ISV or developer building something for selling to others. different needs for them

2 days ago
#

wombling
Chris Paine
@vijayasankarv @esjewett @vlvl @rhirsch @aiazkazi and don’t forget customers with seasonal/fluctuating demand. (Repeating myself, sorry)

2 days ago
#

vijayasankarv
Vijay Vijayasankar
@wombling @esjewett @vlvl @rhirsch @aiazkazi yes agreed – needs to be solved absolutely, either at IaaS level and/or at PaaS level

2 days ago
#

rhirsch
Dick Hirsch
@vijayasankarv @wombling @esjewett @vlvl @aiazkazi 2 sides to consider — shop & platform – both need to support diff models

2 days ago
#

esjewett
Ethan Jewett
@vijayasankarv @vlvl @wombling @rhirsch @aiazkazi Exactly. Much clearer than me :-). I hope SAP covers both. Right now, focus seems on #1.

2 days ago
#

vijayasankarv
Vijay Vijayasankar
@esjewett @vlvl @wombling @rhirsch @aiazkazi which brings up multi tenancy topic . Do u expect it as platform feature or leave it to apps?

2 days ago
#

esjewett
Ethan Jewett
@vijayasankarv @vlvl @wombling @rhirsch @aiazkazi Yeah, very good point. Too complicated for twitter, and I need to sleep 🙂

2 days ago
#

wombling
Chris Paine
@vijayasankarv whilst @esjewett is sleeping 😉 how do you think from a PaaS viewpoint multi-tenancy could be delivered as a feature? (cont)

2 days ago
#

wombling
Chris Paine
@vijayasankarv @esjewett (cont) by building into the security/roles/authorisations of standard IDM solution? Extend to social login?

2 days ago
#

vijayasankarv
Vijay Vijayasankar
@wombling that could be a solution. but fundamentally a principle need to be agreed whether platform needs to even support multitenancy

2 days ago
#

vijayasankarv
Vijay Vijayasankar
@wombling it could also be that apps might want control of how to implement multitenancy without platform dictating it

2 days ago
#

wombling
Chris Paine
@vijayasankarv The worry is leaving it to app developers means potential embarrassment < but at least app developers fault not SAP!

2 days ago
#

vijayasankarv
Vijay Vijayasankar
@wombling it is like C++ and Java 🙂 I didnt like java for a long time thinking it took away my ability to fully control what I am building

2 days ago
#

wombling
Chris Paine
@vijayasankarv Be glad you never had to code Web Dynpro Java then 😉 Or if you did, then I can see yr point very well

2 days ago
#

vijayasankarv
Vijay Vijayasankar
@wombling I was already out of full time dev role by the time WD was widely used – but yes, did a little bit when it came out

2 days ago
#

esjewett
Ethan Jewett
@wombling @vijayasankarv If SAP is going to certify apps as multi-tenant, it’s going to require a manual audit. No pure tech solution.

2 days ago
#

vijayasankarv
Vijay Vijayasankar
@esjewett @wombling rather left field question – do u think a model where apps are not certified by platform provider is feasible ?

2 days ago
#

wombling
Chris Paine
@vijayasankarv @esjewett feasible yes, q: is the value to partner to have SAP logo stamped onto app worth the investment? probably yes

2 days ago
#

esjewett
Ethan Jewett
@vijayasankarv @wombling Sure, but then customer has to trust app dev. You can provide tools, but no way to guarantee data isn’t mixed.

2 days ago
#

vijayasankarv
Vijay Vijayasankar
@esjewett @wombling not a lot a platform provider can really certify beyond some minimum things like ” won’t crash, meets usability reqs” 🙂

1 day ago
#

esjewett
Ethan Jewett
@vijayasankarv @wombling Yup. Multi-tenancy is not offered by any PaaSes as far as I know.

1 day ago
#

wombling
Chris Paine
@vijayasankarv partners will build multi-tenancy solutions (I’m trying now) but social login means can’t leverage IDM solution anyway

2 days ago
#

vijayasankarv
Vijay Vijayasankar
@wombling yes – but do you think a hybrid of social login and traditional MDM can solve it elegantly?

2 days ago
#

wombling
Chris Paine
@vijayasankarv personally I find that too many frameworks complicate solutions rather than making them easier. Eg what happened to SOAP

2 days ago
#

vijayasankarv
Vijay Vijayasankar
@wombling 100% agree – and that is at least partly because very few developers think highly of other developers IMO. Too quick to dismiss 🙂

2 days ago
#

wombling
Chris Paine
@vijayasankarv still, would not be surprised if logic to allow multi-tenancy was delivered as is natural extension of current user mgt

2 days ago
#

esjewett
Ethan Jewett
@vlvl @vijayasankarv @wombling @rhirsch @aiazkazi That’s what I mean by different. Bot sure if it’ll work, or the implications. Interesting.

2 days ago
#

esjewett
Ethan Jewett
@vlvl @vijayasankarv @wombling @rhirsch @aiazkazi Though, SAP seems to take a diff approach to PaaS than other PaaSes. Need to think on it.

2 days ago
#

rhirsch
Dick Hirsch
@wombling related question would be if whether all the plumbing is there to deal with subscriptions #saphanacloud

2 days ago
#

 My thoughts

A couple of days later and my question has be answered on SCN, but I’ve also had a few moments to think about this.

Drinking your own Champagne

Firstly, to my own need for a productive license to some very minimal use of the SAP HANA Cloud.

With the reasonably low price point that SAP is putting on Cloud Partner status, it certainly seems that they are trying to attract small companies to develop content for them. If you add to this, the “we drink our own champagne” marketing message that has been broadcast very well by the ex-CIO there is a obvious marketing proposition.

If companies that signed on as partners for SAP then submitted an application to the SAP Store for resale, they could be allowed to use it productively themselves, they would have an excellent sales pitch “we drink our own champagne”. A limit on the sizing of the used solution might be in order (but probably wouldn’t be an issue with small companies), but it would be very cool for small companies to do this. It would certainly encourage companies like the one I work for to go the extra step of putting the application into the SAP Store. A win for both the developers and SAP.

Annual fixed storage/cpu isn’t elastic, isn’t cloudy for a PaaS

Probably the clearest idea in the thread above is that PaaS shouldn’t be billed annually. Where we are talking SaaS (like Yariv Zur’s SAP HANA Cloud Portal (which is kinda SaaS and PaaS, but I’d argue definitely both)) then there is a different view, but for a PaaS, the beauty of the solution is in its ability to scale up as demand dictates.

I was talking to BCO6181 – Tony de Thomasis’ uni course this evening about the use case where you have a wonderfully capable server that has 10 CPUs running at 1-2% utilisation all year. And you have a policy that no-one gets a pay rise unless they complete their annual performance review. Guess what, on the afternoon before the cut-off, the system is running at 100% capacity and people are complaining about how slow and poor performing it is. In the cloud you shouldn’t have to deal with that. But if you have to buy your cloud compute units annually, you are going to be in exactly the same space.

On the plus side, it looks as if this might be addressed soon. I really hope so, as I see a big potential for SAP HANA Cloud to be the next big thing in enhancing SAP’s cloud SaaS solutions, but if it’s just a glorified hosting arrangement, then it starts to loose some of that PaaS shine.

Thanks to all those who posted their thoughts publicly for me to capture in this blog. I hope you don’t mind me reposting, let me know if you’d like anything redacted.

 

 

 

 

blue sky

Blue sky thinking

The other day I was chatting with someone in 140 character snippets about some of my thoughts on the things that may come to pass. I’d also alluded to my thoughts in much the same space at the “cloud” panel at the SAP Inside Track day at Melbourne (kindly hosted by the nice folks running the Mastering SAP Technologies conference.)

So I thought that rather than just let those words vanish into the air, I’d instead write them down so that sometime in the future someone would have the possibility to point out how wrong I was with absolute certainty.

SaaS solutions and enhancement

I am a strong believer that companies have, and will continue to have a need to build system based processes that are different from company to company. The idea that they will all adopt the same solution because it is “best practice” and the cheapest to use and manage is, to me, very unlikely. Companies will continue to change, vary and improve those processes that they believe bring special value to their business. In many businesses (but certainly not all) I believe that the area that they will need to innovate on is “people”. As repeated by Mark Souter at the recent Mastering SAP HR & Payroll conference and potentially originally by Eddie Barrett from Deloitte (I can’t find any earlier references than 2008!) “The war for talent is over, talent won.” Companies that don’t innovate above and beyond the offerings that SaaS providers give them will lose a competitive edge.

Right now, I’d say that adopting a “vanilla” SaaS HR product (especially one that is currently rated as “market leader” – if you haven’t seen the tweets or don’t get the references then the SuccessFactors team not been doing its job properly 😉 ) would probably give most companies an edge over their competitors, but this will not endure long term.

The thing that has made SAP so well loved by enterprise, has been its ability to allow for enhancement to the vanilla “best of breed” model. Now we all know of cases where the addition to the vanilla was more cod-liver oil than raspberry jam flavoured and the resulting mess has been a support nightmare, but this is not always the case.

Cloud SaaS solutions will naturally evolve (as they aim to replace the on premise models that we have been so used to using and enhancing) such that we will be able to write our own business logic into them. In many cases this will be more configuration than code. And I’m sure that we will see much more care taken with the underlying core of the solution so that it can’t be broken as badly as many on premise solutions have. (After all, one hopes that we’ve learnt something from all these years of wasted effort.) But unless the cloud solutions offered by SAP allow for enhancement, then someone else will build cloud solutions that can be enhanced (natively) and SAP will be either forced to purchase them or lose the enterprise market.

I see SAP HANA Cloud PaaS as the first real step on this journey of SAP SaaS enhancement. I’m fully onboard this train, and I see it accelerating pretty darn quickly.

Running a hosted legacy enterprise solution and offering it as “cloud” is pandering to the uneducated or at best a stop-gap solution.

Ok so this isn’t so much a prediction as a rant. I think any software vendor that offers both a standalone on-premise install of software and a cloud version that are identical in function do not have true cloud software on their hands. The agility of cloud based software should be in its ability to scale, be elastic and to take advantage of the capability to spike to big resources to do amazing things. That sort of agility can’t be used by software that must work within defined hardware limits. If the software is supposed to function equally well in both scenarios, it clearly hasn’t been built to take advantage of the cloud (and never will be able to!)

Current batch process based time and payroll processing is an anachronism and will be replaced.

If SAP aren’t busy building a HANA based payroll with a skunkworks team somewhere, someone else ought to. (If anyone reading this wants me to help form such a team…) With the power of HANA to close to instantly calculate the impacts of any change in circumstances of an employee and resolve their payments for the next n-years, there should be no need for a “pay-run”. Pay would already have been run! Payroll processing would instead be a consolidation/audit process where data entry was checked and variations explained (part of current processing) but the bit where the system was offline for hours to enable the processing would no longer be required. No-SQL has offered this tantalising possibility several years earlier, but the idea of having an eventually consistent payroll by using a map-reduce type approach would have been enough to make most auditors suffer cardiac arrest (much as both ideas might be appealing, you can see why this hasn’t been adopted.) With HANA offering a fully consistent SQL model and the possibility of near real-time evaluation of results there is a new option. Add into this mix the ability to scale up to a cloud sized resources with the ability to elastically scale as demand requires and the possibilities start adding up. Imaging being able to model people movements in your enterprise and have the exact dollar amounts that this would mean. Rostering optimisation would become much less of an art and much more of science.  The space that SAP ceded to third party time management and rostering solutions (such as workbrain) would be ripe for competition, if the payroll and time management functions that currently are batch based became either on-change calculated or close to real-time available.

SAP’s current payroll system is “world-class” I don’t think anything else can do payroll in so many countries. (At least this is what I’ve been told and I haven’t seen convincing arguments otherwise.) But it is based on an architecture that was designed for systems that had less computing power than the desktop computer that I’m writing this blog on (sorry not using my tablet today 😉  ). If SAP want to retain their world class ranking then they need to innovate and rebuild. I’m pretty sure that there is a little team at the very least throwing ideas around on how to achieve that. I predict that by 2017 we will see a HANA based cloud payroll and time management solution(whether built by SAP or a partner, I’m not sure) that will put the current SAP on premise based solutions to shame.

Cloud ERP – same, same but different.

With the possibility to “enhance” the cloud, will come the wider desire for ERPs to be fully cloud based. It doesn’t make sense for any organisation to have to worry about IT infrastructure when cloud offerings are available. I’m just waiting for an analytics firm to write some software which figures out which organisations are still using their own email servers and then starts discounting their share values due to their inherent technical stagnation. (Am I being too harsh here? I’m not sure.) I predict that a HANA based cloud ERP solution will be available for SAP customers by 2020 which will have more functionality than the on-premise solution. In the run up to this, each LOB solution will have a SAP HANA cloud based solution built or purchased and adapted that offers as much functionality as the current on premise model. The work will be in the integration of all the LOB cloud solutions so that they can work as a cohesive ERP.

It is on the work towards a cohesive solution where I believe that SAP spent worthwhile money on SuccessFactors. The SuccessFactors BizX suite including SAP Jam and Employee Central aren’t in anyway built in a homogeneous manner by the same team using the same data model. Rather they are a collection of heterogeneous best-of-breed solutions sewn together to make an even better whole. It is the experience and ability of the SuccessFactors team to take heterogeneous solutions and make them appear as a homogeneous whole to a user that is valuable. (You could argue that they could do a better job, but like Einstein famously said, kissing and driving fast cars at the same time ain’t that easy. As far as I’m concerned it’s a lot nicer than other stuff I’ve had to work with recently.)

Now, I’ve heard some fairly vocal people argue that having a good object based data model that allows for easy cross-component use is essential for any to-be successful SaaS solution. I disagree. It’s how it works for the end user that counts. And the thing is, no matter how beautifully you design your solution, there is always going to be someone that build a little part of it better. If you can’t purchase/partner with that amazing start-up and bring them into your system, well, someone else will. A couple of amazing start-ups later and the most amazing system (from the end user point of view, which is the only one that counts) isn’t the beautifully designed one, it’s the one that can bring all the best bits together, no matter who built them or how they built them or what data model they used.

Blue Sky thinking – starts with BS.

I may be completely wrong with my thoughts. I’ve never considered myself to be a futurist – sounds far too wanky for me. So I’m certainly not very scientific about how I’ve formulated these ideas. But I do think we are on the point where the enterprise software market is turning. And it’s turning towards the cloud.

Or these thoughts could be just a load of wishful BS.