Updates

Author: dukicn

articles/storage/blobs/network-file-system-protocol-performance-benchmark.md

@@ -27,7 +27,7 @@ No matter which method is used, it's always important to understand other potent
 - testing with `cp` command copies one or more files from source to the destination storage service and measuring the time it takes to fully finish the operation. This command performs buffered, not direct IO and depends on buffer sizes, operating system, threading model, etc. On the other hand, some real-world applications behave in similar way and sometimes represent a good use case.
 - second often used command is `dd`. Command is single threaded and in large scale bandwidth testing, results are limited by the speed of a single CPU core. It's possible to run multiple commands at the same time and assign them to different cores, but that complicates the testing and aggregating results. It's also much simpler to run than some of the performance benchmarking tools.
 
-**Using performance benchmark tools** represents synthetic performance testing that is common in comparing different storage services. Tools are properly designed to utilize available client resources to maximize the storage throughput. Most of the tools are configurable and allow mimicking real-world applications, at least the simpler ones. Mimicking real-world applications requires detail information on application behavior. 
+**Using performance benchmark tools** represents synthetic performance testing that is common in comparing different storage services. Tools are properly designed to utilize available client resources to maximize the storage throughput. Most of the tools are configurable and allow mimicking real-world applications, at least the simpler ones. Mimicking real-world applications requires detail information on application behavior and understanding their storage patterns. 
 
 **Using real-world application** is always the best method as it measures performance for real-world workloads that users are running on top of storage service. However, this method is often not practical as it requires replica of the production environment and end-users to generate proper load on the system. Some applications do have a load generation capability and should be used for performance benchmarking.