Microsoft Outlook and the Enterprise
August 1, 2008
If you are curious about the impact of Gmail on Microsoft Outlook mail, you will want to read “Breaking Free of Outlook” by Bernard Lunn in the August 1, 2008, ReadWriteWeb here. The core of this essay is about the writer’s shift from a fat client like Outlook to the Google cloud-based email service Gmail.
Mr. Lunn points out two immediate advantages: almost no spam and he could locate messages. Now Gmail is not flawless. Certain file extensions are not supported. The non hierarchical approach to organizing email takes a bit of adaptation or it did for this addled goose. I know about indexing, but I had to change my behavior in order to make full use of the Google “tags” notion. Note: Diversity blog has some useful comments about Gmail’s disintermediating utility and small business here.
Mr. Nunn points to Microsoft’s concern about Exchange, Microsoft’s mail server. He writes:
This game is now clearly all about economies of scale on those giant server farms, so we are likely to see email server hosting consolidate down to a handful of companies in the next few years. This is the normal and expected lifecycle for a commodity market such as email serving.
Nail hit on its head. Mr. Nunn hits the *key* point in the contention between Google and Microsoft. Let me offer several observations:
- Microsoft is building data centers at a ferocious clip. More information here. The assumption is that these data centers will run as fast or faster than Google’s data centers. My research suggests that Google has a performance advantage ranging from 2X to 12X. If these data are accurate, Microsoft will have to build twice as many data centers as Google to match Google’s throughput. Any other order of magnitude means that Microsoft cannot catch up to Google just building data centers with available technology.
- Microsoft is spending money on brand name gear. Google, by contrast, buys commodity hardware. Google also automates many routine set up and maintenance functions. The result is a cost savings for Google. Google buys a commodity server at the going rate for a dual processor, dual core pizza box with four gigabytes of RAM and dual SATA drives. Our price work ups suggest that Google can buy one server for about about $700 to $800 based on ArnoldIT’s analysis of the commodity components. A comparable device from Hewlett Packard can costs anywhere from $1,100 up to $8,000 or more, depending on the configuration, maintenance plan, and other facets of the purchase agreement. The $600 difference is meaningless for large companies but when scaled across 10,000 or more servers per data center, the price difference begins to mount up. Other savings kick in so Google spends less to deliver its performance.
What this means is that Google can scale while spending less on hardware and achieve a significant performance advantage. If these data are accurate, it seems clear to me that Microsoft is going to have to spend more money than Google to match Google’s performance levels. This means that Microsoft may need 1.5 to 2.0 or more servers than Google to deliver comparable throughput. If true, Microsoft will have to spend even more money to reach parity with Google.
Now, the killer in this scaling arms race. Google continues to push forward in its infrastructure performance. By working around known bottlenecks such as data writes and reads, Google does not have to buy exotic and expensive hardware to resolve these choke points. That’s another cost and performance advantage which ripples through set up, support, and maintenance tasks for these devices.
When I read Mr. Nunn’s comment about “economies of scale”, I thought of Google’s price performance advantage. If my data are correct, the cost gaps between Google and Microsoft are so significant, that I can chop these estimates in half. Microsoft still has to spend more than Google to match Google’s performance. Because Google continues to make incremental improvements, Microsoft will need plenty of money to deal with email and other type of cloud computing scale issues.
Microsoft can achieve parity with Google–at a price. A dip in Windows or server license revenues can impair Microsoft’s ability to continue to spend for data centers to deal with the types of scale Mr. Nunn references.
The metaphor that comes to my mind is that of a well funded Formula 1 racing team trying to catch up with a special purpose jet powered car designed for the Bonneville salt flats. The cost of reworking the Formula 1 infrastructure makes the exercise an extremely time consuming, difficult, and expensive one. The jet powered vehicle uses a commodity engine, racing tires, and a 20-something bold enough to fire up the engine and head down the track. But those Formula 1 team members look classy in their matching jump suits if that’s any consolation.
Stephen Arnold, August 1, 2008