Failsafe: When asked about the possible risk of having one important part of possibly dozens or hundreds of files in one chunk, I was told that several failsafe mechanisms are in place that perform file system health checks on the specific area that the deduplicated data gets written to. Also, if your server accesses data very frequently (they wouldn’t share a number), it gets automatically duplicated again to avoid possible risk.
And, by the way, this very same technique applies to network transfers (think VDI!) and even local RAM. Windows Server 8 saves local memory by finding duplicate information and getting rid of it.
ChckDsk Ultimate: Live disk fixing
After decades of putting up with Chkdsk, which essentially takes your servers offline for hours in case of a hard drive disaster or crash, Microsoft came up with a major revamp of the infamous disk checker tool. ChkDsk offers (finally!) an online scan and corruption logging mechanism. It marks defective clusters and files during runtime and cleans them later using a technique called "Spot Repair," which shows itself both through the UI (Server Manager, Action Center) and PowerShell.
The "Spot Verifier Service," which is part of the Windows 8 client as well, attempts to repair disk issues on-the-fly. If that's not possible, maybe due to a more critical file system error, admins have the chance to attempt and perform an instant repair using the the "spotfix" parameter or schedule a chkdsk for later.
If possible, chkdsk simply eliminates the errors that the "Online scan" feature logged, which takes only a few seconds or minutes compared to the hours that you'd have to wait on Windows Server 2008 R2.
Network performance and reliability
Being a server OS, Microsoft invested quite heavily in speeding up network performance and robustness. D-VMQ (Dynamic Virtual Machine Queue) helps avoid CPU bottlenecks in high-bandwidth situations by effectively alligning the network traffic to processor cores -- you'll end up using only a fraction of the original CPU performance under high bandwidth usage. Second, the overall priorization of traffic has been improved, with the admin being in full charge of dynamically reserving bandwidth for particular traffic. Second, the new "NIC Teaming" feature allows you to combine adapters and essentially scale network performance with every NIC that gets added (e.g., get two 10 GB NICs and essential drive 20 GB down the lines). What's really interesting about this is the fact that you can combine NICs from different vendors (e.g. Broadcom or Intel).
Follow Sandro on Google+
