Soft Restart as a Mechanism to Lower Risk

SoftRestartasaMechanismtoLowerRisk()

Windows Server 2016 is just around the corner from release. There has been a good amount of information about features such as Nano Server and containers. One feature that would appear to not only be an incremental improvement, but a very welcome addition is the ability to perform a soft restart.

Traditionally, when a physical server performs a reboot, many things happen before services are up and an administrator can log back in. The server BIOS does some level of initialization, at times skipping some of the extended memory tests. Sometimes the baseboard management controller takes a moment to inventory the system. Any host bus adapters will spin up drives and describe arrays, as well as load BIOS extensions. Network adapters will generally load BIOS extensions as well. This can often add minutes to the boot process. I know I’ve felt anxious while watching a constant ping on more than one occasion.

The soft restart feature in Windows Server 2016 is meant to speed up this entire process. So far, this is implemented in two varieties. One can reboot from the Windows command prompt with shutdown.exe /r /soft and the other from PowerShell with Restart-Computer -Soft. Regardless of which way you choose to do it, what you get is a restart of the operating system only, without the hardware initialization step. We have determined that this can save anywhere from 25-45% of the boot time.

We did some real world testing to understand where the savings are most dramatic. Workstation hardware showed only a negligible increase in boot speed due to not having any special HBAs or NICs to initialize. Virtual machines showed even less of a return due to the way the underlying hardware is virtualized. Physical servers are where this feature is truly valuable. On the limited set of servers tested, we were able to get in the range of the 25-45% savings.

Not much information is available from Microsoft just yet. We believe this feature is designed to reduce the risk of failures during general patching and upgrades on clusters that would momentarily not have N+1 redundancy. On physical systems that also are not capable of a high availability setup, this decreases downtime during periods of planned maintenance. We’re sure that Microsoft will release more information as general availability status for Windows Server 2016 approaches.

Windows Server 2016 Blog Series

Innovative Integration is creating a whole series about Windows Server 2016 leading up to the September launch. To read other articles from this series, click here

About Jeff Peters

Expert in many aspects of the Microsoft Product Line, including SharePoint, Active Directory, Server, Exchange and Office 365, SQL, Clustering, Operations Manager, Lync/Skype for Business, and Configuration Manager. Currently holds MCSE (since 2005) and MCITP. Experience with designing and implementing Windows server environments, corporate e-mail, and directory services. Experience with Operations Management and Configuration Management software. Exceptional expertise with portal design, custom development, CRM, XML datatypes, and Forms Services.

Leave a Reply

Innovative Integration can help you optimize your IT infrastructure. Request a Consultation