Appliances offer more than 'On Premise' less than SaaS

One of my favorite bloggers, Bob at SmoothSpan has a post (almost one month old now, but its taken me that long to think through it) which discusses the question Are appliances SaaS?

A growing number of vendors are delivering a service platform where their software is either deployed to the customer
1. In a Hardware package which is plug and play in customer network.
2. A Virtual Machine which can be downloaded and installed by the customer.

These appliances contain all the application components needed to run the service application including the DBMS.

Appliances are definitely an improvement over a traditional On premise solution, but there are a number of key support and operational challenges that SaaS handles well and Appliances do not.

1. Appliances tend to have better Version Control and Management for Upgrades than On Premise solutions. Vendors can push updates out to their entire network of installed appliances, which means there is no version mess and support burden of handling different products.

But there is a question mark about the reliability of this process when dealing with large numbers of appliances. I can guarantee that if you send an update to enough appliances, that a number of these updates will fail.

This could be caused by the usual risk factors:

1. Network outage on Customer side during. Less likely to happen in an established hosted provider like Opsource
2. If the software update has database schema changes, something could crash during this process. If you are updating thousands of databases this could very well happen.
3. Security and Permissions have been changed.
4. Hardware failure

On the other hand, one of Bob's respondents stated a major benefit of a Virtual Appliance is that it gives the client the opportunity to test the update before applying to production and even goes so far as to say that this is preferred at times over the risk of a single update in a true SaaS application. I would suggest that the opposite is true. In one of my earlier blog posts I discuss the huge problem 'on premise' vendors face with versioning and upgrades. The same holds true for virtual appliances albeit with less risk of the OS environment changing. Customers will test at their own pace, and more often than not the decision to upgrade will be delayed sometimes indefinitely. You therefore lose some important agile benefits of a SaaS system. Customers do not regularly get updates and miss out on new features, the SaaS vendor spends more time on supporting, developing and testing for different versions with updates having to handle different environment scenarios which has a direct impact on service delivery, and more time is spent working with the customer through the QA phase and working through potential objections to the upgrade.

There is no doubt that Appliances are superior alternatives to On Premise, in fact an integration appliance should be your first choice, if you have a mix of SaaS and On premise systems which require integration from behind the firewall. A Good example of this is Cast Iron Systems which has appliances for SaaS systems Salesforce.com and Rightnow.

The agility of SaaS vs. the risk of On Premise

I read an article today from PC World titled SAP uninterested in enterprise SaaS

This is a very typical response from a traditional enterprise software vendor who underestimates the pervasiveness and stickability (I like that word a lot better than traction) of SaaS. Lets look at a couple of quotes from this article.

The ERP giant was "not so interested in 'drop-in' services for three or four users, he said. "We want to run the company, we want to run the business, we don't want to just support some services for some users in the company."

This simply indicates the inflexibility and non agile nature of On Premise vendors. Of course they don't want to support 3 or 4 user systems. The cost of sales and support is just too high for them. On premise vendors struggle with both providing customers with new features regularly and version control which Bob at Smoothspan blogs so effectively about.

Back in my On Premise Days we struggled with these exact same problems which haven't changed today.. We had a major product release once a year, the problem was and is how did you deploy that release to all your customers. In the end you don't.

The whole on premise product release process is non agile in nature because of the risks involved. You can't bring out a product release more than once a year because you have to test every permutation of what could go wrong with each customer's installation and build upgrade procedures for all previous versions in existence. You then announce to the customer that there is a new release with a whole bunch of new features and it will cost them 1 million in services to upgrade. They look at the new release, and then say "thanks but no thanks". They look at the risk of something going wrong in the upgrade and say its too risky and too expensive.

In the end you find yourself using your latest release for new customers only and you are stuck with a version quagmire. Its a vicious cycle as each new release then has to cope with yet another older version upgrade.


John Rowell from Opsource also posts about this issue of SaaS vs. on Premise

The beauty of SaaS is obvious, one instance of the application to upgrade and all customers benefit from new features. Because of this, SaaS becomes very agile. You can bring out cool new features once a month because you don't have the baggage of versioning and handling different customer environments.

SaaS also is a whole different philosophy when you begin looking at user counts.
SaaS doesn't care if you are 25000 users or 3 to 4 users. It can handle it all. There is no "sorry you are too small for us". SaaS can't handle large user count systems? Somebody better tell Salesforce.com that, earlier this year they closed a 25000 user deal.

"They would prefer a quick win, that could be some CRM functions on demand," he said.
They might later want to bring in on-premises software, he said. "That would take the risk and cost down."

It was suggested in this article that SaaS was seen by enterprise customers as a short term fix before going to a long term on premise solution.

Once again I will speak from personal experience. We had one enterprise prospect, it was down to us and a traditional on premise vendor, we won the deal because our competitor quoted an 18 month project, with a 7 figure cost, plus annual maintenance fees. We offered a SaaS solution and said we could roll out immediately and would release new features once a month. 6 months later we had one thousand users in multiple countries all using the same instance as all other customers. Consolidation of country data was painless, (can you imagine what this would have been like with an on premise system) and customer feature requests were being delivered monthly with our entire client base benefiting from these enhancements.

Can you imagine a customer in this scenario saying lets implement this SaaS system just for the short term and then go On Premise in the long term? Not Likely...