Netmail Blog

Microsoft Exchange Architectural Evolution – Cloud Messaging Services and the Evolution of Choice

Posted by Judi Kohn

Find me on:

Jun 2, 2014 9:31:00 AM

Technology. Most of us would agree that it is a key part of the world around us. In the field of technology, the first sixty plus years have seen spectacular progress in all aspects of the domain. Ever shrinking transistors have enabled the production of increasingly complex and cost-effective microchips, a scenario that has driven the surge in technological design. This fits with Moore’s Law, the observation (not a physical law) that the number of transistors on integrated circuits doubles approximately every two years. 

This observation is frequently used in the semiconductor industry to set research and development targets. In 2005, this growth rate was expected to continue until at least 2015, but the 2010 update to the International Technology Roadmap for Semiconductors predicted that the rate would slow at the end of 2013 when transistor counts and densities are to double only every three years.

A colleague of mine, Eyad Saheb, told me that he spent an afternoon thinking about changes and advances in technology and he noted his thoughts on the current state of Microsoft’s Exchange architecture. Since I thought that it was a thoughtful and insightful reflection on the topic at hand, I have reproduced it below.


History repeating is well-known phenomenon. George Santayana is famously quoted as saying, “Those who cannot remember the past are condemned to repeat it.” However, whereas some repetition is a result of not learning from past mistakes, sometimes it’s circumstance that leads you back down the same path.

Strangely enough, good examples can be witnessed in the realm of technology. It may surprise, given the relentless march of technology, that it would ever return to what existed prior (the 5 ¼-inch floppy disk is not likely to resurface anytime soon), but technology also changes the circumstances of the world around it. Consider, for example, Microsoft’s Exchange architecture. Two decades ago it was a simple, all-in-one mail server. But subsequent iterations saw it partitioned into functional roles, separate and specialized. Yet returning to present day Exchange 2013, one sees the roles collapsing back into one another,1 supplemented by Microsoft’s strong recommendation that all roles be deployed on each server. The modern day multi-role server is effectively the monolithic mail server of the past. Sure, servers can now be clustered to act as a single system, but the top-level architecture has come 360 degrees, nonetheless.

Similar things are happening on Exchange’s bottom-level architecture. Early generations of the architecture would accept whatever enterprise disks the server had available to it, but subsequent versions greatly benefited from high-throughput SANs. This led organizations to become accustomed to hosting mail storage on very expensive and sophisticated hardware. Fast forward to Exchange 2013, and years of persistent database optimizations have dropped hard disk I/O by approximately 94% compared to its early predecessors. Microsoft is now pushing a JBOD2 architecture (to the dismay of admins everywhere that spent half their IT budget on a shiny new SAN).

Looking beyond Exchange, to a more general client-server architecture, we can find further examples of the past in disguise. As computing power grew, early day thin clients connected to mainframes eventually gave way to thick clients, but at the cost of performance: network performance lagged. Today, multi-gigabit networks deliver next-generation services like SaaS, PaaS, and IaaS, ushering in another era of thin client computing (Amazon’s Kindle Fire being a fine example).

So where does this leave us? If the trend is to repeat what already existed in the past, does it mean we haven’t actually progressed? Clearly a rhetorical question, as the benefits of technological progress cannot be denied. The difference between what was then and what is now, is choice. Choice is what turns the cycle into a spiral, turning 360 degrees while still moving forward from where it began. In the past, architecture was bound into thin client or thick client routes (for example) because of technological constraints. Today, we have what we need to let us choose which is better for the goal at hand.

Let’s return to Exchange to illustrate the point. Cloud services like Office 365 now permit organizations to offload their messaging infrastructure to Microsoft. This is not unlike a thick client to thin client transition. But just because technology now permits this, does it make it the right choice for everybody?

The decision to hand over years of sensitive company communication should not be decided by crunching numbers. Risk is a very real, and unquantifiable, consideration. What if no-one’s watching it, or watching it a little too closely? What if there’s a security breach? Can you be sure it will be detected, and are you sure you would be notified? What if there’s an issue with your Cloud account, and how important is that issue for the service provider relative to their other clients’ issues? Will resolving your issue be a priority? These types of questions haunt CIOs across the country as the allure of cloud services becomes ever more enticing. Each success story of a smooth migration is tainted with tales of another client that decided to pull it all back in-house (at significant cost). But this is where progress has provided us choice. The decision is no longer forced on us by the technology, and better yet, it’s not even a binary decision anymore. You CAN take both. Move some mailboxes to the cloud, just not all mailboxes. Move some data cloud, just not all data. Solutions such as Microsoft’s hybrid Exchange, coupled with Netmail Archive, allow you to draw the line at a point you feel comfortable. Is the sanctity of the CFO’s mailbox keeping you up at night? Keep that one in-house. What if a security breach allowed access to all emails? Keep only a few weeks online, shuffle the rest to on-premise archives. Admins have all the tools they need to break from the historic cycles, and customize a path that’s right for them.

How do you feel about the current trend towards Cloud services for your organization’s messaging needs? Are you comfortable with it? Is all your data stored in the cloud or is it on-premise? Is a Hybrid solution the best for your needs? The choice belongs to you…

1Actually, I have it on good pretty good authority that the only reason the CAS role is still separate is because Office 365 needs it that way.

2Just a Bunch Of Disks

Topics: Email storage, Email Management, Office 365, Netmail Archive, Microsoft Exchange 2013, Cloud


Subscribe to Email Updates

Recent Posts

Follow Me