You might have seen Frank Anderson’s take on the performance impact of Office 2013. This article inspired Project VRC to perform additional investigations in Office's performance impact and best practices. As a result we are publishing today a new whitepaper: Project Virtual Reality Check - Phase VI - Impact of Microsoft Office on VDI.
The reality today is that Microsoft Office is by far the most used application suite in the corporate environment. The Office suite is almost always deployed and used in virtual desktop infrastructures. But what is the performance impact of upgrading Microsoft Office?
This whitepaper focuses on the performance and capacity impact of the three most recent Office versions in VDI running Windows 7. Microsoft Office 2007, Office 2010 and Office 2013 are reviewed with both x86 and x64 versions of Windows and Microsoft Office.
This research led Project VRC to the following conclusions:
Comparing both Microsoft Office 2010 and 2013 with Office 2007 there is a negligible capacity impact of 1% with Office 2010, but there is a very substantial difference of 20% with Office 2013. As a result, upgrading to Office 2013 requires 20% more VDI capacity in comparison to Office 2007 and Office 2010.
Office 2013 will consistently use more CPU and much more memory. For instance, Office 2010 will consume 26% more memory than Office 2007, which is surely nothing to laugh at. Office 2013, on the other hand, uses a rather stunning 272% (averaged) more memory on an application level.
Project VRC also tested possible performance optimizations for 2013 such as “animations off” and “hardware graphics acceleration off”. These optimizations did not make a positive performance difference.
Reviewing Office 2010 x86 and x64 running on Windows 7 x64 it is possible to conclude that the x64 architecture had no significant impact on VSImax, when CPU is the only limiting factor and enough memory and IO capacity is available. However, running x64 version of both Windows and Office will have substantial impact on storage IOPS and memory footprint. For instance, Microsoft Office 2010 x64 running on Windows 7 x64 will consume 32% more memory compared to Office 2010 x86 running on Windows 7 x86.
Indexing (search) was also reviewed to understand how it affects VDI capacity. Disabling indexing is considered a performance best-practice, however, it is a highly-appreciated feature by Office Users. If enough IO capacity is available, enabling indexing has only a 3% impact. If storage allows a potential 30% increase in write IO’s, it difficult to recommend disabling search, since it is such a critical feature to user acceptance.
Many organizations consider an upgrade to Office 2013 for its functional benefits. However, upgrading to Microsoft Office 2013 will have a significant user capacity impact within VDI. The impact of Office 2013 on CPU, memory and disk IO is considerably higher than Office 2007 and 2010. Currently, there are no obvious tuning best practices available to lower Office 2013’s performance impact. It is therefore highly recommended to evaluate the performance impact of Office 2013 in your own environment before it is deployed.
Read this and more in the latest paper which is available for download at www.ProjectVRC.com (free, registration required). Ruben and I would like to thank Ryan Bijkerk who set up and executed so many tests, collected all the data and wrote this excellent paper.