Far fewer writes have to go to storage
The real important thing here, the takeaway point is: a fairly high percentage of writes are no longer having to go to storage. It’s not that the systems are not doing as much work; they are just now doing it smarter, and more logically organized.
Larger, more sequential I/Os
They are doing it in larger, more sequential I/Os than haphazard, random, tiny, fractured I/Os.
More I/Os = More Time
Each of the I/Os takes time to complete. So, the more I/Os you do, the longer it’s going to take for that transaction to complete its task.
If we’re able to eliminate 30-40, nearly 50% of the writes from having to go to storage, an interesting thing happens here: however you write the data is how Windows is going to read it back. Technically the IntelliWrite functionality is providing a benefit on those initial writes and any subsequent reads.
Free, idle, and unused memory
Now we can take that a step further by identifying and leveraging free, idle, and unused memory on that particular Windows guest VM. Or if it’s a physical Windows Server, that’s fine too. Or workstation. But a Windows system that’s no longer or not using all of that available memory, we can temporarily borrow a portion of that memory to act as a repository for our repetitive read cache, where we identify the same logical blocks that are being read over and over and over again.
Windows can retrieve data directly from memory
It doesn’t make sense to keep kicking those out and going all the way to storage. If you have memory that we can put a copy of that data and now have Windows retrieve the data directly from memory, it no longer has to send that read I/O request to storage.
12 to 15 times faster than flash or SSD
And, because that data is being recycled or being read multiple times, it’s getting satisfied at the speed of memory. Memory-to-memory data transfers are like 12 to 15 times faster than going to flash or SSD.
Storage can now be more efficient
And your backend storage is getting the benefit that it didn’t have to waste its precious time even worrying about or processing that request. The storage system can now focus more of its attention on the I/Os that do have to go to storage, those writes and non-repetitive reads.
Resources:
Article: I/Os Are Not Created Equal – Random I/O versus Sequential I/O
Case Study: Telestream solves timeouts and slows on I/O intensive applications
Article: Myriad of Windows Performance Problems Traced to a Single Source
Article: How To Get The Most Out Of Your Flash Storage Or Move To Cloud
Video: The Two I/O Fallacies Surrounding IOPS and I/O Response Time