Believe it or not, 12 substantial Windows performance problems that can cause the most frustration and chew up valuable time can be directly traced to a single source. In this article, we’re going to show you how. First, let’s briefly touch on and describe each issue.
1. Slow Application Performance This is an issue that many of us are familiar with. A company is running a large application such as EMR/EHR or ERP that the entire enterprise depends on, and users end up waiting endlessly for data. Or a sales team is operating on a CRM application, and speaking with prospects; waiting for data in such a scenario is not helpful to getting a sale made. It could be an LMS, used for the vital administration of educational programs. Other applications such as SharePoint, MS Exchange, VDI, POS and even legacy and proprietary apps all suffer from this same malady. This makes your phone ring or support tickets blow-up with user complaints.
2. Application Crashes This one is especially annoying, and just brings everything to a dead stop. Freezes and crashes are the biggest headaches of IT, there is nothing worse than angry users. As an example, a customer service representative is on the phone with an important customer, and while looking at the customer’s data, the screen freezes up. The application has crashed. And, oh yes, this will affect others accessing that application, too. When this happens, often a user will yell out, “What’s wrong with the computer?!” But of course, it’s not the computer. We’ll get to that at the end.
3. Missing SLAs SLAs are the delivery backbone of many companies. Service quality and availability are service aspects written into contracts, and if they’re not met, it not only means lost income, it can also mean lost business and clients. This is especially true today in a SaaS environment, in which a client can simply pull the plug and go to another provider. A primary cause of missed SLAs is slow performance. Yet again, it traces to the same source as these others.
4. Slow Data Transfer Rates There are many reasons for data transfer, including backups to other locations, and importing data to new locations. But they all boil down to transferring a large amount of data from one place to another. When transfer rates are slow, it means waiting. And waiting. And waiting. This Windows performance problem eats up system as well as staff resources. Slow data transfer rates are traceable to this same source.
5. SQL Query Timeouts and Latency Today, many enterprises survive on data, which means they’re also living and dying on database queries. When a query is originated, the process through which the query was made will be waiting until the query is satisfied. The longer the wait (latency), the more cost is eaten up in terms of time and resources. If a timeout occurs, that means that the query must be started again. This, of course, can mean a serious delay.
6. SQL Deadlocks This phenomenon occurs when two or more processes are waiting for the same resource. Each process is then waiting for the other process to complete before continuing. On the user end, SQL deadlocks produce the same result as timeouts: endless waiting.
7. SQL Server 15-Second Warnings An I/O request should complete within milliseconds. The 15-second warning that SQL Server has been waiting for longer than 15 seconds for an I/O request to complete indicates a serious performance problem—once again traceable to the same issue.
8. You Upgrade Hardware…but Performance Still Slow Many seem to think that the best (and maybe the only) way to solve performance problems is to upgrade hardware. But what happens when you upgrade hardware, and performance is still sluggish? This is simply a very expensive way to indicate that you have “solved” the wrong problem. Yes, performance was an issue, but the reason behind it was not hardware related. Yes, you guessed it: the cause is the same as all of these other problems.
9. Slow SSD Read/write Speed Following along on the above scenario, many companies install SSDs to improve performance—and given the substantial performance difference between SSDs and HDDs, that performance difference should be drastic. But what happens when the read/write speed to SSDs is still slow? Yes, you’re still suffering from the same problem.
10. Storage Performance Problems Storage has become very advanced today, with sophisticated solutions designed to improve storage performance. But just as with the issues described above, it often happens that the performance problems you’re experiencing with storage are not due to the hardware…but to the same cause as the rest of these issues.
11. Slow Server Performance This is the generally sluggish performance phenomenon, the causes of which can be tough to trace down. For that reason, many don’t try—they just decide that hardware must be upgraded: new servers, new storage, perhaps even a new network. In that slow server performance is most often rooted in the same cause as all of these other issues, though, you might want to give that a try first. Or at the very least, you can buy much less hardware and save that budget!
12. VM Density and Consolidation Issues VM consolidation is the action of consolidating several VMs into one physical server. Virtual machine density means the quantity of VMs being run from a single physical host; the higher the VM density is, the more efficient the system may be. Both VM consolidation and VM density contain the same inherent performance problem as each of these other scenarios and may be preventing you from loading more VMs onto a single host.
The Basic Problem
All of these Windows performance problems that cost you peace of mind can be traced back storage I/O efficiencies.
As great as virtualization has been great for server efficiency, one of the biggest downsides to virtualization is that it adds complexity to the data path – otherwise known as the I/O blender effect that mixes and randomizes I/O streams.
There are 2 severe I/O inefficiencies causing this.
First, is caused by the behavior of the Windows file system. It will tend to break up writes into separate storage I/Os and send each I/O packet down to the storage layer separately and this causes I/O characteristics that are much smaller, more fractured, more random than they need to be – this along with the I/O Blender effect noted above is the perfect trifecta for bad storage performance.
This is a “death by a thousand cuts” scenario that is like pouring molasses on your systems – everything is running, but not running nearly as fast as it could.
You could opt to throw more hardware at the problem, but this is expensive and disruptive and can be premature – it is much better to tune what you already own to get the performance you should be.
Second, is storage I/O contention. This happens when you have multiple systems all sharing the same storage resource.
Windows is breaking up that I/O profile into a much smaller, more fractured, more random I/O profile than it needs to be. If you just clean that up on one VM then all of the data from that one VM to the host is all streamlined, but then you have all the data from neighbor VMs that are still noisy and causing contention.
As you can see, your performance is not only penalized once, but twice by storage I/O efficiencies. This means systems process workloads about 50% slower than they should on the typical Windows server because far more I/O is needed to process any given workload. This has been found to be to be the cause of a host of Windows performance problems such as those mentioned earlier.
In order to achieve an ideal performance level from your hardware infrastructure, you want large, clean, and contiguous read and write I/Os from all sources, eliminating the I/O blender effect.
Larger, cleaner, sequential I/Os result in fewer I/Os to process and thus faster data transfer rates for peak performance. In such a case, you can have 1G of data, but instead of transferring it in 100,000 I/Os, you can accomplish it in 70,000, or less.
The next factor is reading and writing I/Os sequentially, instead of randomly. When dealing with storage, you’ll find that sequential I/Os always out-perform random I/Os on hard disk drives, SSDs and flash storage.
These 3 factors work together to transform the nature of the I/O to improve performance:
- Larger I/O
- Sequential I/O
- Less I/O The overall effect is that the OS workload is reduced, because there are fewer I/Os to process, and they are occurring sequentially.
This is the solution brought into effect by the DymaxIO fast data software:
- Fewer I/Os, because they are larger
- Sequential I/Os
- Read I/O served from memory DymaxIO accomplishes these improvements through proprietary technology that optimizes and streamlines with both reads and writes.
Write performance: IntelliWrite® patented technology eliminates small, fractured I/Os caused by Windows splitting files into multiple write operations. DymaxIO enforces large, clean, contiguous writes for more payload with every I/O operation.
Read performance: IntelliMemory® patented technology reduces read I/Os from storage by caching hot data server-side. Reads are cached right at VM level from otherwise-idle, available DRAM. Not only does this enormously decrease the I/O latency time, but also decreases the I/O traffic to the storage unit, thus freeing up the storage bandwidth for other work.
Because of these substantial improvements, DymaxIO is able to regularly provide 30 to 40 percent faster data transfer speeds, eliminating a myriad of Windows performance problems.
Are your servers good candidates for DymaxIO ? Find out quickly and easily without investing a lot of time – Download the FREE Condusiv I/O Assessment Tool. This free tool will:
- Analyze data across 11 performance metrics
- Easily identify systems suffering from performance issues
- Graphs display averages and peaks for each hour
DymaxIO improves the performance and reliability of Windows systems.