Friday, January 28, 2005

On demand computing: the rebirth of service bureaus

Application service providers (ASP), which provide application software hosting, were supposed be the "next big thing" of the Internet boom. But, around 2001, ASPs fell into disfavor, along with all things dot-com.

But now that the hype is gone, ASPs are quietly gaining ground, mostly under the latest buzzword, "on demand computing." The latest evidence is that IBM is buying Corio, one of the major remaining independent ASPs. It's a small deal for IBM--the price tag was only $182 million. But it indicates that IBM, which spends a lot of time thinking about the future of IT, thinks that on demand computing is more than the latest marketing slogan.

Back to the future
Whether you refer to it as on demand computing, or the ASP model, it's really the return to a very old concept in business computing: service bureaus. Service bureaus started in the late 1950s when computers were very expensive and beyond the reach of all but the largest companies. So in order to expand the market and make computers available to smaller companies, the big computer vendors, such as IBM, Burroughs, Univac, and Control Data, set up service bureaus to sell computer time on a shared basis. In other words, time-sharing.

Today, businesses are faced with a problem that is the same--only different. The problem is not that computing hardware is too expensive--those prices have been dropping every year to the point the laptop I am using at this moment to write this article has more computing power than the IBM mainframe that I programmed twenty years ago. The problem today is that software packages, especially enterprise applications, such as ERP and CRM, are too expensive and too difficult to administer for small companies. Smaller companies might be able to afford the hardware, but most can't afford the IT staff and infrastructure that it takes to support a major enterprise application. Therefore, a service bureau, or time-sharing approach, needs to be applied, not to hardware but to software. We need service bureaus for software.

Service bureaus, time-sharing, ASPs, and on demand computing are pretty much all the same concept. The customer doesn't run the system, and may not even own the system. The vendor runs the system, and the customer pays for it as he uses it.

For IBM, on demand computing is the closing of a full circle. IBM was forced to abandon time-sharing as a business in 1973 when it settled an antitrust lawsuit by selling its Service Bureau Corporation subsidiary to Control Data Corporation (CDC). Now in offering application hosting, and buying Corio, it's fully back in the time-sharing business.

Software vendors offering on demand services
Among the major enterprise software vendors, Oracle has been particularly aggressive in promoting its ASP service, called "Oracle OnDemand," which allows small companies to deploy Oracle's entire E-Business Suite, with Oracle hosting the system in its own data center for a maximum of $150 per user per month. The customer still has to pay Oracle's software license fee, but the hosting service relieves the customer of having to provide the infrastructure and staff to support the system. Other major vendors, such as SAP, offer similar programs.

The CRM vendors in particular seem to be having the most success in promoting the ASP model. Vendors such as have been gaining traction, as small companies and even some large companies find that the ASP model makes a lot more sense in supporting sales reps as they roam the field. is a pure-play ASP system, designed specifically for the hosted model, and the monthly per-user charge, which starts at $65 per month, includes everything--there's no up-front software license fee. Faced with the success of, CRM market leader Siebel last year acquired hosted CRM provider Upshot and is now offering it as its own Siebel CRM OnDemand, going head to head against at a comparable $70 per user per month.

In addition to the, there are other pure-play ASP systems. NetSuite, an ASP that built its own enterprise system offering from the ground up to take advantage of the ASP model, continues to expand its product footprint, and now offers complete ERP (NetERP) and CRM (NetCRM) functionality, on top of its accounting system formerly known as NetLedger. Unlike Oracle OnDemand, NetSuite does not require an upfront license purchase, but can be purchased simply on a per-user, per month, basis.

Advantages of on demand computing
The ASP model is a good fit for small companies that do not have the IT staff or infrastructure in-house to manage information systems. Implementation is typically quicker than with systems deployed in-house, and the in-house support requirements are fewer. The ASP is responsible for all necessary software upgrades and system maintenance. The ASP also relieves the customer of having to worry about IT infrastructure requirements, such as security, backups, and disaster recovery. The ASP model also makes good sense for systems that support employees that are frequently on the road--applications such as sales force automation, expense reporting, and field service support are proving to be good choices to deploy on a hosted basis.

My own consulting firm, Strativa, is a good example of a company that benefits from using an ASP. We use a professional services automation system on a hosted basis that takes care of all our CRM, time and expense tracking, and billing requirements. Even though we do technology consulting, we'd rather have our staff spend their time delivering consulting services than maintaining an internal information system. Therefore, the ASP model is perfect for us. For a flat monthly fee, it gives our consultants instant access to the system anywhere, any time, through a Web browser, and it requires zero internal technical staff.

There is one other application that is well suited to the on demand model, and that's e-commerce systems. Because such systems can experience sudden and huge variation in transaction volume, it makes sense to host such systems under an on demand model that can flex to provide the computing power needed to accommodate such spikes. For example, if you host your own Web storefront and you're selling a product that suddenly gets favorable press in the Wall Street, your sudden good fortune can bring down your Web storefront. Not a good thing for customer service. Similarly, few small companies can afford the IT security and recovery measures that need to be in place to provide secure e-commerce. Therefore, small and mid-sized companies ought to consider the on demand model for any Internet-based application that are critical to the business. Even larger companies should consider this approach.

Disadvantages of on demand computing
Where the ASP model does not make a lot of sense, however, is in large companies that have multiple systems that must be integrated with each other or with custom systems written internally. Someday, with widespread adoption of a service oriented architecture (SOA) by software vendors, it might become easier for customers to integrate in-house systems with ASP-hosted systems. But today, if you have a mix of systems from different vendors, or a combination of packages and custom software, it's better to locate such systems within the same corporate data center.

Update, Jan. 30. Darrell Miller has posted a long comment to this post, disagreeing with my view:
The problem is that we are trying to avoid the root cause and find a workaround solution. As you stated "enterprise applications, such as ERP and CRM, are too expensive and too difficult to administer for small companies". This is the problem that we should be attempting to fix. CRM apps and ERP apps should not be difficult to administer.
Darrell makes some good points. Read his whole comment, then post your thoughts. What do you think?

Related posts
ASPs making a quiet comeback
Siebel loses $59M and responds by going on a shopping spree
Mid-market may be sweet for CRM ASPs


Darrel Miller said...

Hi Frank,

I do not see this model working for small companies either. Here are some problems that I see:

- Any situation where there is a significant amount of data entry to do, web applications fall horribly short.
- Integration with other desktop applications, e.g. Outlook, Excel, Crystal Reports is difficult.
- You are now vulnerable to connectivity outages. As we have seen in the past twelve months, this can happen to the biggest of providers. Small companies are not likely to have much influence over a big service provider.
- What would it do to a business if the service provider went out of business? ISP's come and go at an incredible rate, why would these "on-demand" suppliers be any different.
- What happens as soon as there are customizations to the hosted s/w?

Regardless of these issues, the problem as I see it is more fundamental. The problem is that we are trying to avoid the root cause and find a workaround solution. As you stated "enterprise applications, such as ERP and CRM, are too expensive and too difficult to administer for small companies". This is the problem that we should be attempting to fix. CRM apps and ERP apps should not be difficult to administer. Sure they are going to take a while to get implemented, but those are business issues.

From a technology standpoint there is no reason why ERP/CRM installs should not be a point and click affair. Years ago enterprise databases required a guru to install. However, Microsoft decided to focus on making SQL Server easy to install and administer and today the entire thing can be installed without asking the user a single question.

Once an ERP system is installed for a small company, IT changes should be extremely infrequent. The administration effort should be insignificant. If it isn't then in my opinion that is a problem with the software design.

We have been through this whole situation once before when everyone decided we should move from thick/rich/fat clients to web applications. The primary reasoning was that web applications were easier to deploy and were accessible remotely. Who cares that they are a completely crippled application platform. Six years later developers are starting to push back towards the cleverly named smart clients. Why, because the problems of deployment and remoting are completely solvable problems and the platform doesn't suck.

As far as IT staff are concerned, small companies need more help supporting their desktops than servers. Sure the servers take a bit longer to set up initially, but once they are running, they rarely need to be touched. Also, small businesses don't need a cluster of high availablity servers. They can buy low cost and easy to manage off the shelf servers. However, the client desktops get messed up all of the time due to viruses, spyware, crappy software installations, etc. Small businesses usually don't have locked down environments like you will find in big businesses. The "OnDemand" guys are not solving those problems.

Personally, I think "OnDemand" is going to go the same way as ASP. All the big players are trying to scale down to reach the little guys, but they don't get it. Small business owners want more control, not less. In the meanwhile I think our focus should be on identifying why enterprise class systems are difficult to administer and we should start to address those problems with the people who can do something about it, i.e. the developers, instead of hiring high priced administrator to hide the problems from us.

Darrel Miller

Darrel Miller said...

I just wanted to add one exception to previous comments. I do not think small businesses should attempt to host their own outward facing servers such as web sites and ecommerce sites. As Frank stated the load on those sites can change quickly and it would be wasteful for small businesses to buy excessive capacity just in case. The other reason is for security. Keeping a machine secure that is exposed on the internet is a task for an IT professional.