A Googler Explains Why Microsoft's Cloud Services Crashed Today

Google data centre uses seawaterA Google data centre.

Photo: Google

Microsoft’s current competitor to Google Apps, known as BPOS, suffered a three-hour outage this morning.During the outage, Microsoft explained via Twitter that the next version of its cloud services for business, Office 365, has been rebuilt from the ground up to be more reliable.

Why did Microsoft have to do this?

Google Apps leader Rajen Sheth has a theory.

Basically, Microsoft has been TALKING about cloud services for the last several years, but a lot of its biggest customers have probably been using hosted servers instead. When it comes to true cloud computing, Microsoft is still getting its footing.

There’s a subtle difference here that’s important to understand. Microsoft kicked off BPOS in 2007 as a dedicated service. Basically, each customer got Exchange (and other software) running on its own server hardware. Customers would save money by outsourcing IT costs to Microsoft, but the technology was no different than running Exchange on site.

Sheth explains that Gmail and Google Apps were designed and launched more than five years ago as a multitenant solution. That means that that the service itself and underlying data is automatically distributed among thousands of physical machines.

Google argues this is a much better way of doing things:

  • Reliability. If a hosted server goes down, customers might notice a problem — at the very least, performance will suffer while they’re switched over to the backup server. With a multitenant solution, the system automatically adjusts, putting applications only where there’s enough capacity and swapping data and services around as needed. Only massive system-wide failures will be noticed.
  • Cost. With a hosted server, each server still has to be managed — hardware has to be fixed, patches have to be applied, and so on. As a result, it’s usually more expensive to run, and those costs are passed along to customers.
  • Updates. Hosted servers are updated the same way on-premise servers are — each physical server has to be taken offline and the software has to be updated. This means updates are usually pretty rare — in the case of Microsoft, they’re usually timed to big new releases of the equivalent on-premises product. (In other words, Exchange 2010 comes out, and at some point the BPOS Dedicated customers get the option to upgrade to Exchange 2010. Eventually, they’re forced to move.) With a multitenant solution, new features can be added regularly and users will “just get them” — just as happens with consumer online services. 

Microsoft later launched a multitenant version of BPOS in 2008. It’s called BPOS Standard.

But that service has apparently been plagued by outages — including the one this morning — and even Microsoft seems to be admitting it wasn’t ready for prime time. Why else redesign it from scratch? (The likely reason: Microsoft’s own server products were not originally designed to be multitenant, so it has taken some time for the company to rebuild the underlying software.)

This will begin change next week with Office 365.

In fact, Microsoft is so confident that the multitenant version of Office 365 will be able to run at scale, it has raised the floor for dedicated (hosted) service from 5,000 to 30,000 seats. In other words, only the very largest Microsoft customers will even be able to get dedicated hosted servers from Microsoft. Everybody else — including the mid-sized enterprises that are Microsoft’s bread and butter — will be steered toward the multitenant version of Office 365.

In other words, Microsoft is in perfect agreement with Google — multitenant is a better solution. It’s just taken a lot longer for Microsoft to get there.

Business Insider Emails & Alerts

Site highlights each day to your inbox.

Follow Business Insider Australia on Facebook, Twitter, LinkedIn, and Instagram.