Cloud, SaaS and SoSaas - Same old Software, as a Service

[Bromcom-PR]

The NIST definition defines the Cloud Software as a Service in similar terms "The capability provided to the consumer is to use the provider’s applications running on a cloud infrastructure" - how the provider provides elasticity (or even whether it does so efficiently) isn't a consideration. The NIST does say "Note: Cloud software takes full advantage of the cloud paradigm by being service oriented with a focus on statelessness, low coupling, modularity, and semantic interoperability." but doesn't define why the "Cloud Paradigm" delivers any of this within its own definition of Cloud - I suspect what they wanted to say was "HTML REST" rather than Cloud. I'm guessing that everyone here's expectation of a cloud delivered MIS would be HTML + REST (which would be a sensible assumption).

 

Extract from UK Government Cloud Strategy Paper: https://www.gov.uk/government/publications/government-cloud-strategy

 

The US National Institute of Standards and Technology‟s (NIST) definition of cloud computing is the most widely adopted one, and has been adopted for G-Cloud; it states that:

 

“Cloud computing is a model for enabling convenient, on-demand network access to a shared pool of configurable computing resources (e.g. networks, servers, storage, applications, and services) that can be rapidly provisioned and released with minimal management effort or cloud provider interaction.”

 

This cloud model promotes availability and iscomposed of five essential characteristics (On-demand self-service, Broadnetwork access, Resource pooling, Rapid elasticity, Measured Service); threeservice models (Cloud Software as a Service (SaaS), Cloud Platform as a Service(PaaS), Cloud Infrastructure as a Service (IaaS)); and, four deployment models(Private cloud, Community cloud, Public cloud, Hybrid cloud). Key enablingtechnologies include: (1) fast wide-area networks, (2) powerful, inexpensiveserver computers, and (3) high-performance virtualization for commodity hardware.

 

SaaS is part of cloud definition and hence SaaS application needs to meet NIST definition of cloud. As to key the subject matter in this thread, Cloud, SaaS vs SoSaaS - client-server applications when hosted as SoSaaS, fail to meet cloud definition in its severe shortcoming in characteristics to share executable code and RAM working storage; client-server application behave as originally designed in an environment that (a workstation) where such resource sharing would have never been needed to be considered. This non-resource-sharing characteristic curtails scalability of such application. Therefore client-server application is not viable as a commercial proposition for scaling beyond minimal to moderate size users against native web-application. The comparison RAM storage requirement is up to 50-fold more for client-server application compare to same application written as web application. You need to add also the extra cooling, electricity supply and of course rack space needed!

Edited February 24, 2014 by Bromcom-PR

[PhillipHamlyn]

Hi Bromcom-PR,

 

I think you make the point very well; given all the information you provide from the UK Government Cloud paper, the UK G-Cloud still allows the provision of SosaaS to be presented for rental to government customers. I recognise that SosaS is nowhere near as efficient (if centrally hosted - not just centrally managed and provisioned) than an equivalent zero-state application design, but this measurement of efficiency would be evident to potential customers because it would be reflected in the service charge to the customer - inefficient centrally hosted applications will be more expensive to rent than efficient centrally hosted ones. The G-Cloud does not insist however that any application is centrally hosted; just centrally and efficiently provisioned, and available at a published price.

 

So; apart from the cost benefits (gained by efficient application design) that are expected to be delivered by a SaaS solution, whats the group's weighted shopping list of features (interoperability has been mentioned by mikecampbell, but presumably there are others such as device independence etc) that are not strictly anything to do with SaaS but are expected to be delivered via a SaaS application ? What does SaaS mean to the community other than the raw "centrally provisioned, and not perpetually owned" definition ?

[GREED]

@PhillipHamlyn can you please describe who you are and what you do a bit please? Your points are really interesting and would be good to know from what perspective you come from with your points.

[Bromcom-PR]

Hi Bromcom-PR,

 

The G-Cloud does not insist however that any application is centrally hosted; just centrally and efficiently provisioned, and available at a published price.

 

Hi Philip,

 

If you ask G-Cloud Team, they will confirm readily that any cloud/SaaS offering that relies on application software that falls outside NIST defintion of cloud will be be allowed on G-Cloud.

 

G-Cloud Team confirmed that client-server applications fall outside NIST cloud definition and will not be allowed on G-Cloud store. If you have any different formal information please share with us.

[matt40k]

I'm guessing that everyone here's expectation of a cloud delivered MIS would be HTML + REST (which would be a sensible assumption).

 

Good point, I often use CentraStage as a good example of Cloud services - however my biggest complain is the lack of an publically open REST API - it's all well and good that they can record a huge wealth of information but it's not achieving it's full potential by being able to fully integrate with other data sources. Being able to integrate cloud service with other services including on-prem is critical. Rackspace and Microsoft both see this.

 

The comparison RAM storage requirement is up to 50-fold more for client-server application compare to same application written as web application. You need to add also the extra cooling, electricity supply and of course rack space needed!

 

95% of all stats are made up. Could you please cite this otherwise it's just another made up stat. I'm not arguing a multi-tenant\cloud optimized requires less resources than a single-tenant\non-cloud which isn't optimized - that would be silly. I'm just saying quote your facts otherwise it'll be like when people say Capita annual maintenance is £20k per year for a small primary (it isn't it's for a secondary being forced to re-licensed).

 

@PhillipHamlyn can you please describe who you are and what you do a bit please? Your points are really interesting and would be good to know from what perspective you come from with your points.

 

He be Capita. (I suspect he updated his signature a little while ago).

[GREED]

You and your facts Matt!

[Bromcom-PR]

95% of all stats are made up. Could you please cite this otherwise it's just another made up stat. I'm not arguing a multi-tenant\cloud optimized requires less resources than a single-tenant\non-cloud which isn't optimized - that would be silly. I'm just saying quote your facts otherwise it'll be like when people say Capita annual maintenance is £20k per year for a small primary (it isn't it's for a secondary being forced to re-licensed).

 

 

Hi matt40k,

 

The comparison RAM storage requirement is up to 50-fold more for client-server application is not from stats. They are results of tests. You can repeat reasonable easily if you want to check as follows:

 

Simulate any client-server based MIS application (or any other application for that matter) running in hosted environment. Measure the memory is being been taken each time you sign in via client application as an extra user or even open a new session for the same user. You will see that client exe will capture 250 MByte RAM for the exe code and working RAM storage. This is because the exe behaves just as it would have behaved in workstation/PC in a local network connected to the server - client exe is not aware of multi-tenancy demands in cloud and hasn't been designed for the exe code to be shared - ie in computing terms it is non re-entrant. Hence failing to meet NIST cloud definition at fundamental level in sharing resource.

 

If you repeat the same tests with a web-based application, you will find that because code is shared (re-entrant), each user takes only 5 MByte RAM and this is for extra users own working storage area. Web backend and client has been designed from outset as a resource to be shared - meeting NIST cloud definition. That is why you can massively scale up and host thousands/millions of users with web applications economically.

 

I hope the above helps.

Edited February 25, 2014 by Bromcom-PR
Formatting quote

[vikpaw]

@PhillipHamlyn can you please describe who you are and what you do a bit please? Your points are really interesting and would be good to know from what perspective you come from with your points.

Defo Capita : http://www.edugeek.net/forums/mis-systems/58928-sims-intouch-have-you-signed-up-2.html#post567853

I always thought he was a system architect or someone quite high up in the dev team, as the help with technical sims error messages has been right on the nail.

 

Loving this discussion by the way.

 

Out of interest, would people consider SLG SAAS or close to it? Since it was written to be held centrally and should respond better, quicker etc. I don't mean in terms of it being scalable on-demand, paid for by unit etc. just in terms of the difference between client-server model to cloud based. Or is it too simplistic to just consider browser based access as being more cloudy? I feel that quite often people want browser based, and they think of that as cloud, but is that necessary fulfilling the definitions...

[PhillipHamlyn]

Greed,

 

We've met before - I presented at the EduGeek Conference with my colleague a few years ago on the subject of the movement to "industrial computing" clouds. My post is Chief Architect at Capita SIMS. I'm interested in this thread because the definition of SaaS is slippery and despite some formally proposed definitions, it definitely means different things to different people; for some it means "cheap, reliable software that I dont own" (the minimalist point of view) all they way to very specific technologies; I need to make sure I am keeping up with the communities aspirations in this regard, not just what the formal definitions are stating.

 

As an example; would the community be satisified that a "app" delivered solution which stored its data in the Cloud but was locally installed (like all iOS Apps) was a valid SaaS application, or does the local install invalidate it, irrespective of the cost-free nature of the installation ? Where do people perceive the highest value of an SaaS application to be - purely financial (i.e. its cheaper than an on-premises solution) or is there a need for it to conform to some other principals (and if so, what is that need being driven from) ?

[matt40k]

You will see that client exe will capture 250 MByte RAM for the exe code and working RAM storage

 

What client exe?

 

If you repeat the same tests with a web-based application, you will find that because code is shared (re-entrant), each user takes only 5 MByte RAM and this is for extra users own working storage area

 

What web-based application?

 

Sorry if I'm being picking but until some independant person\company publishes a report showing their findings quoting numbers like 250MB of RAM vs 5MB of RAM is total rubbish. I've just opened SIMS and done a few things and it's sitting at 24,316KB, I'm sure if I allocate more RAM it'll eat it, or if I spent the entire day in SIMS it'll grow but it's still a way off from your 256,000KB you've quoted.

 

Like I said, I'm not arguing that optimized vs non-optimized won't help, I completely agree, sticking something like SIMS .net in the cloud would be silly when you compare it to something like your MIS that has been designed for the cloud from the ground up. I'm just saying stick to the actual hard facts. You're not having to waste resources running multiple copies of the OS for example. Rather than quoting unscientific numbers, it doesn't help. They get quoted as fact then business cases fall part at the first huddle because the facts aren't actual facts.

[matt40k]

@PhillipHamlyn the biggest benefit of cloud is the "big data" idea. Yet no-one has yet mastered it in the MIS market.

[PhillipHamlyn]

Bromcom-PR,

 

I take your point but some of the offerings on G-cloud are virtual desktop environments with full desktop versions of Microsoft Office (not just browser) available to the end user (vendor is InsightCloud) . I'm not sure this falls within NIST or not but if a remote desktop version of an SosaaS is deemed "Cloud" by G-Cloud then I think its a useful extra definition of "what makes a cloud application Cloud'y". In this example the vendor provides user interfaces "Browser, Other, Client Side Virtual Application" which does indicate they support some level of client side deployment (even if its App Streaming or somesuch).

 

Personally I'd love to know how those chaps offering virtual desktops full of Office productivity software avoid the inefficiencies of hosted desktop software that you described in your earlier post - in general its very difficult to achieve elasticity for remote application hosting. Time to get hunting for some technical background on their offering I think :-)

[Bromcom-PR]

 

What client exe?

 

 

What web-based application?

 

Client exe: SIMS.net or Facility CMIS will do.

 

Web-based application: You are welcome to use Bromcom Cloud MIS or any other web based MIS.

 

It will not be practical to guide you on this forum to simulate a 'hosted' environment', adding in users and memory measurements etc... But in essence you need to measure "free memory" in the 'hosted environment' after initial a cold start then after each time a user is log.

 

PM me and I will provide you with the test report/procedure followed and the results.

[PhillipHamlyn]

Matt40k,

 

Does that mean data cross schools, or just the data within a school ? I can see that, for example, Fisher Family Trust provide a "Big Data Insight" service to the education market, is this the kind of thing you mean ?

[GREED]

Greed,

 

We've met before - I presented at the EduGeek Conference with my colleague a few years ago on the subject of the movement to "industrial computing" clouds. My post is Chief Architect at Capita SIMS. I'm interested in this thread because the definition of SaaS is slippery and despite some formally proposed definitions, it definitely means different things to different people; for some it means "cheap, reliable software that I dont own" (the minimalist point of view) all they way to very specific technologies; I need to make sure I am keeping up with the communities aspirations in this regard, not just what the formal definitions are stating.

 

Thanks Phillip and my apologies. I have no recollection, are you sure I was at that one?

[pwainewright]

@PhillipHamlyn Thank you for joining in, interesting comments.

 

I must say, though, the Gartner definition of SaaS you cite is so impossibly broad as to be totally useless. If all it takes is a subscription contract and remote management, then my family's Microsoft Office, for which I pay an annual subscription and which is kept up-to-date by Office Update, counts as a SaaS application - which patently it is not. The SaaS version is Office 365 although it is only partially SaaS since the client (unless you use the web version) still runs on the desktop.

 

I believe the NIST definition often gets misinterpreted because it describes three separate cumulative layers: IaaS, PaaS and SaaS. While it's possible to host a client-server application on IaaS and deliver it from the cloud, that doesn't make it SaaS. To be described as a cloud application per the NIST definition, an application has to include core architectural features in the definition such as resource pooling and elasticity. Otherwise it's just SoSaaS sitting on IaaS, and hugely inefficient/unwieldy.

 

I'm interested to see integration and big data cited as advantages of cloud SaaS - I think those are good points. I know I'm a bit of an interloper here but in the interests of stimulating discussion, here's a list of cloud-native SaaS benefits that I included in a recent article on client-server vs cloud SaaS:

◾It’s quick and easy to get started and to add or reduce resources due to on-demand, elastic provisioning and highly automated operations.

◾A high-performance, globally connected infrastructure comes built in as standard.

◾The application always remains bang up-to-date with the latest functionality, updated automatically to add new features without disruption to existing settings.

◾Intelligent resource usage across a common, demand-pooled application infrastructure allows for cost-effective service delivery.

◾Collective testing and feedback on the shared infrastructure ensures robust security and reliable operation.

◾A highly composable service-oriented architecture provides a flexible, fault tolerant platform with extensive scope for continuous innovation at all layers.

◾The ecosystem of users and partners can easily share add-ons and templates that all run on the same shared, homogenous operational instance.

◾Open, cloud-ready APIs and high-bandwidth connectivity make it easy to plug in external resources and functionality.

[matt40k]

@Bromcom-PR - I look forward to your nice public report... hint hint.

@PhillipHamlyn - Think bigger.

[GREED]

@PhillipHamlyn the biggest benefit of cloud is the "big data" idea. Yet no-one has yet mastered it in the MIS market.

 

I'd like to think Aspen came close on that front Matt Cross country potential...

 

Interesting how the DE will make big data accessable

[PhillipHamlyn]

pwainewright,

 

Thanks for your comments, I can see that the concept of "software as a service" means much more to you in its architectural design than the raw definition implies. You are pulling in concepts from open API systems, software design stuctures, low operational cost (as opposed to low price), shared infrastructure (not neccessarily, in my opinion, a good thing on a business critical system), and high bandwidth connectivity. I'm not pushing back on your definition of SaaS but is a lot more extensive a shopping list than I'd expected to be loaded onto it. Very much food for thought.

[matt40k]

I'd like to think Aspen came close on that front Matt Cross country potential...

 

Yup. Gutted about that. Don't get me wrong, far from perfect, but it's defo something new.

 

One can only hope.

[Bromcom-PR]

@Bromcom-PR - I look forward to your nice public report... hint hint.

 

We have been sharing this report on commercial confidential basis with bono fide interested/relevant parties. Release it public??? On a forum?? Hmmm - we'll need to think about it. Meanwhile our offer to PM me to release to you with restriction on circulation is still on the table. However you will be welcome to share with the forum / community as to your copies and conclusions you draw from the report.

[Bromcom-PR]

@Bromcom-PR - I look forward to your nice public report... hint hint.

 

We have now provided matt40 report and also guidance how step by step is able repeat the tests. For running the tests matt40 appears to have access to SIMS. We are happy to provide access to Bromcom MIS.

 

So it is down to matt40 to enlighten the forum on the facts of 50-fold memory wastage if client-server MIS is hosted rather than web-based MIS.

 

Press Release: Laboratory tests prove up to 50-fold saving for Web-Based School MIS over legacy Windows ‘fat client’

 

Meanwhile we have this press release that provides essence of the report: http://www.bromcom.com/press/pdf/130112_Press.pdf - so the findings have been in public domain for sometime.

[Bromcom-PR]

We have now provided matt40 report and also guidance how step by step is able repeat the tests. For running the tests matt40 appears to have access to SIMS. We are happy to provide access to Bromcom MIS.

 

So it is down to matt40 to enlighten the forum on the facts of 50-fold memory wastage if client-server MIS is hosted rather than web-based MIS.

 

Press Release: Laboratory tests prove up to 50-fold saving for Web-Based School MIS over legacy Windows ‘fat client’

 

Meanwhile we have this press release that provides essence of the report: http://www.bromcom.com/press/pdf/130112_Press.pdf - so the findings have been in public domain for sometime.

 

Upon his request, we now also provided GREED with the report.

 

So we now have two active members of EduGeek MIS forum on the job and hopefully soon both will enlighten the forum their comments and take on the facts presented in the report re: up to 50-fold memory wastage if client-server MIS is hosted rather than web-based MIS.

[PhilNeal]

Could I have a copy?!

[matt40k]

I wouldn't bother @PhilNeal. If it was worth while, they'd publish it publicly.

 

First, thanks @Bromcom-PR for the kindly sharing the document with me and offering access to Bromcom to replicate the tests.

 

I think we got a bit muddled up about the objective of this topic, as I understand it we all agree - A web based MIS system written from the ground up to be cloud optimized, like Bromcom, is better at running as a hosted service rather then a traditional client\server application, like SIMS .net. I would be pretty sure we'll all 100% on this, even @PhilNeal is with us on this point. So great, why are we using SIMS again?

 

Now we go down hill it what I like to refer to as a Bromcom special - simply put, over selling.

 

So let's start with - http://www.bromcom.com/press/pdf/130112_Press.pdf - MY issue, is your statements like:

 

50-fold reduction in RAM needed per user for the web based application

 

The crux of how you got this number is you ran SIMS on RDS, done a few things the look at resource manager and noted the memory used. My issue is if say I was Cambridgeshire LA and I put in a request to change SIMS for Bromcom, I can't use this as a bases of cost saves. It's laughable. You're comparing apples to oranges. There is no talk about what the requires are on the end-client are because I've yet to find a web browser that doesn't eat RAM. There is no talk about Disc IO, I mean I could allocate SIMS 20MB, I'm sure I could get it working in that and most importantly no actual response times of how long stuff takes to actually perform, I mean it could take 10 mins to save a single attendance far as I know!

 

OK, let's put that to the side for now and move on, lets say it's all correct. Massive savings, any Bromcom quote should look like Capita quote only with the decimal point in a place that makes bursars everywhere happy.

 

So lets go back to the PDF above: "Legacy desktop – locally hosted £1,500 per primary"

 

Sorry, that's £300,000 for a 300 pupil primary?! Seriously? Just on admin side? Really? So let's say that's 15 classrooms of 20 students, say another 3 office staff, bursar, head teachers, so 20 machines, SIMS is, say £1k, let's throw on a few addons, so that's say another £3k. Say another £10k for a server - so that's £34k, oh let's slap some Capita consultant on top, what's that these days? £1k a day? Hell lets have 14 days - so that £48k, still another £252k to spend! Hey how much is SLG these days?

 

At this point I'm starting to question Bromcom alot more. Now they offered me access to Bromcom, by access they mean over Go-To-Meeting type thing or by visiting and using a laptop. So I can't actually compare like for like or do any actual real testing - ie point a monitoring system at it then run over a long period of time on identical hardware. It's all right playing the Cloud card, but us tax payers end up paying for the resources when you boil down to it. I'm basically not bothering to test it unless I can get both systems on my own hardware. It's not worth my time running a bias test scenario and it certainly isn't worth public money. It then makes me wonder why they wouldn't want to reveal the application? I mean I'm happy to sign a NDA. Which then makes thing, has it been PEN tested?

 

Amazing how a bit of over selling with some over optimistic "facts" has turned me from a lover to a hater lol.

 

 

(PEN testing is penetration testing, there are different types, the type I am referring to is application penetration testing. Whilst other testing and standards refer to how you patch, handle data, secure your firewall and blah blah blah, application pen testing is where a security expert inspects the application and basically tries to hack in. They don't care if what they find causes your project to over run or break. The idea is your security is only good as your weakest point. A really good case of this not occurring is Tesco's - Troy Hunt: The Tesco hack – here’s how it (probably) happened)