In August, Microsoft massively increased the capacity limits on its Azure Backup data protection offering. Now, the Redmond, Wash., software and cloud services provider is going for speed.
"Azure Backup now leverages USN journal (Update Sequence Number Journal) technology in Windows to track files that have changed between consecutive backups," announced Giridhar Mosay, a Microsoft Cloud and Enterprise program manager, in a company blog post. "USN keeps track of changes to files and directories on the volume that helps in identifying changed files quickly, enabling faster backups."
Now, backing up Windows PC, Server, SharePoint, Exchange and SQL can take half the time, said Mosay. "We've seen up to a 50 percent reduction of backup times in volumes with two million files utilizing this optimization."
Azure Backup now features more efficient metadata handling, reducing the size of the cache required for the service to run. "Previously, Azure Backup required cache space of 15 percent of volume size being backed up to Azure," a requirement that was generally acceptable for smaller volumes but "became prohibitive with volumes greater than 10TB," said Mosay.
"We introduced a new algorithm to compute this metadata that utilizes far less disk space," he revealed. "In our internal tests with large volumes, we are now seeing less than 5 percent cache space requirement which is a 3X improvement. To that effect, we are updating our requirement for cache space accordingly to be less than 5 percent of the size of data being backed up."
Microsoft also increased the maximum number of recovery points from 366 to 9,999. Other improvements include more forgiving timeouts to ensure that "long running jobs complete reliably," Mosay said. The company also decoupled backup data cataloging from the mainstream data upload process, allowing the service to manage cloud backups more efficiently, he added.
Also this week, Microsoft released a newer version of Azure Data Lake Tools for Visual Studio that allows developers to run U-SQL scripts locally on their own machines.
"This is a powerful capability when you are developing new code or any time you want to test and debug in tight loops," said Omid Afnan, principal program manager of Microsoft Big Data Platforms, in a Dec. 9 blog post. "The local run feature supports all U-SQL scripts including ones with C# user code. You can reference files, ADL [Azure Data Lake] tables and other metadata just as you would during cloud execution."
Introduced by Microsoft in September, U-SQL is a new query language that combines the benefits of SQL with the expressive power of a developer's code.
"Code-based solutions offer great power, but require significant investments to master, while SQL-based tools make it easy to get started but are difficult to extend," said T. K. "Ranga" Rengarajan, corporate vice president of Microsoft's Data Platform, in a Sept. 18 announcement. "We've faced the same problems inside Microsoft and that's why we introduced, U-SQL, a new query language that unifies the ease of use of SQL with the expressive power of C#."