Saturday, January 31, 2009

Multi-tenancy is Better for You - the Customer

Lately, the topic of multi-tenancy and single tenancy has again come up for discussion. A leading on-premise vendor recently argued in favor of single tenancy by saying:

11 comments:

Raj said...

Your analogy of planes and motorcycles is really stupid. The plane going down doesn't make headline because it is only one instance. It takes down 300 people with it at the same time. If 300 motorcycles crashed at the same time killing all its riders, it will be a front page news.

So the question is that of scale. How much wider impact a single instance going down can cause?

From a perspective of a user, a single tenant software is definitely better. Only he gets affected.

However, the only point where you can argue about the viability of a multi-tenant software is that of the economies of scale. The customer gets it cheaper because the costs are shared. If cost is not of concern in the particular case, I would go with single-tenancy.

Anshu Sharma said...

Raj - Actually you prove my point. If 500 motorcycles crash killing the rider in a month, one at a time - there is no news event. But if 1 plane with 500 people on it crashes, its big news. Clearly, the bikes are no safer than the plane. And newsworthiness is not same as safety.

You say - "From a perspective of a user, a single tenant software is definitely better. Only he gets affected." I disagree. How does others being up help you? Actually it hurts you because the vendor may not find you going down to be enough of a high priority issue.

You may want to re-read my post before commenting.

Andra said...

I actually agree with Raj here. The analogy does not really make much sense. If 500 motorcyclists crash in one day, yes it will create news. Hence a single tenant software would not cause a major havoc for the critical customer.

Anshu Sharma said...

I don't understand Andra. The single tenant system going down will not cause a major havoc for the software provider. For the customer, if he is down - he is down (and alone in his woes).

KRD said...

Let me join the above discussion in comments.
You are thinking from POV of a software provider, not a customer. The customer does not want to get on a front page news, does he? If the plane goes down or the motorcycle crashes, the customer is dead anyways. The last thing you want to do for a dead guy is to put him on the front page. With single tenancy, he is atleast alone in his woes and not making the front page.

Anshu Sharma said...

Actually - the reverse. If the Boeing plane of an airline crashes, its Boeing & the airline that get into trouble. This keeps them honest. (Similarly, software vendors that service multiple customers try to keep the systems running to avoid problems.) This implies greater effort on part of the service provider (Boeing, airline, multi-tenant software provider) - to build, manage and operate systems that fail the least.

On the other hand, if your software is installed one unit at a time and it goes down - the only person impacted is that customer. The vendor (software provider or motorcycle manufacturer) has no skin in the game.

Anshu Sharma said...

In my previous comment, I should have said - the vendor has less skin in the game. Clearly, all vendors - motorcycle & planes, single-tenant and multi-tenant - want to do well by customers. Its just that in some cases the incentives make them try harder.

Anonymous said...

Anshu - this post is not well thought out mate.

So 500 users sit in a single plane. Consider this plane to be a single stack of hardware, software etc. supporting 500 lives.

You have many such planes in the sky. One goes down - it does not impact 50,000 people in the sky at any given time - does it? Again in SaaS multi-tenancy - it is not one giant system shared with many customers but risk mitigation where many users are located to a mini-stack. Not a single-tenant.

Your motorcycle example - i have words for it. It seems like a malware or spyware attacked 500 computers simultaneously and took them down. Doesn't seem SaaSy at all to me.

Anonymous said...

really bad analogy, planes and motorcylces and how can you suggest that because its a plane you as an airline manufacturer care more about death than the motorcyle manufacturers....?! Death is not the issue here its flexibility and choice...

lets look at a better auto example buses and cars or in the sfdc case 1 big bus.

the fact is that if the bus crashes everyone on the bus stops. If a car crashes everyone else in their cars keep driving. Because the bus has to cater for everyone it bigger and slower, takes a generic route to accommodate all passengers.

The drivers of cars can go where they please, stop when they need to etc. Cars are more flexible there are more options, and last I checked there is a lot more choice in which car I buy. Cant see me buying a bus anytime soon.

FYI I am a car driver and dont take the bus.

Anonymous said...

You got your a.. handed to you on this post. Didn't you? Hahahahaha!

Vijay Jagdale said...

Raj, Andrea et al: Seriously, how can you be so dense (and rude) as not get such a simple analogy? Statistically, airplanes are thousands of times safer than a motorcycle as a mode of transportation. However, in the very rare instance of an airplane mishap, the poor airline has to deal with the nightmare of the Press onslaught and questions of what went wrong. Motorcycles otoh, kill thousands or people everyday, (one at a time), but rarely get any media attention!

If I needed hosting services, from a strictly cost/benefit/reliability analysis, it would behoove me to use a multi-tenant solution, because it usually has robust architecture, better security/backup/uptime, beefier hardware, benefits of load-sharing, overall cheaper solutions, and so on...

Agreed, when the multi-tenant hardware fails, it brings down all the tenants on that platform, but strictly from a statistical standpoint, it is far rarer incident then the single tenant solution, and a more robust, and cost effective alternative.