DataCore’s MaxParallel for Windows Server application acceleration software has just been released for non-workload specific environments. After success with MaxParallel for SQL Server during late 2017 and early 2018, DataCore have now released a version of the acceleration software that is not tied to a specific type of database or other workload.
So what is MaxParallel for Windows Server, and how could it help you in your Data Center? And how and why should you accurately test the software to give meaningful results that deliver real benefits in your own environment?
What is it?
MaxParallel for Windows Server is the third generation of DataCore’s MaxParallel software. Versions 1 and 2 were specifically targeted at Microsoft SQL Server environments, and would accelerate only those workloads. The current release of the software, R3 has been opened up to allow your other Windows based applications to benefit from the parallel processing capabilities it offers. Your Exchange Server, Oracle, MySQL and other similar high throughput workloads, could now also benefit to gain a much needed performance boost, and possibly allowing you to get more work done with fewer resources, alongside a potential to save on licensing costs.
We’ve created MaxParallel for Windows Server to address one key challenge – what many people fail to realize is that there exists a bottleneck in the OS, in that it passes I/O between the CPU and the disk subsystem in a single threaded, interrupt based manner. This obviously slows things down. You could try to fix this in a number of ways, for example by installing expensive all-flash storage either in the systems themselves or as an all-flash array, but that still won’t remove the bottleneck in the OS. It’s true that the I/O that is in flight to or from the storage might arrive at its destination faster, but there will only ever be a single I/O in flight. Enter MaxParallel!
MaxParallel as a Windows Service
MaxParallel installs into the Windows Operating System, and runs as a service. Once installed, it works by using idle time on the system’s CPU cores to push more work through to the backend disk subsystem. There is no modification required to the workload application (database etc.) or additional hardware required. A single reboot during your existing maintenance window, and then it’s ready to go.
Optimal Testing of MaxParallel
But how should we go about testing the MaxParallel implementation in a meaningful fashion? We need to ensure that we are creating a workload that is as close to “real life” as possible. Ideally we would have a testbed that mimics the production environment as closely as possible, but this is not always possible. There are however, tools that can be used to generate this meaningful test data, and a number of ideas are given here.
- For databases, there are tools both within and without the database server itself that can be used to test effectively. In newer versions of SQL Server, you can use Distributed Replay, which captures real life workload and allows you to replay it against the test database(s) to gauge the performance levels that can be attained with and without MaxParallel for Windows Server.
- Another tool for databases is hammerDB, a freeware tool with a wide community involvement that allows simulation of both TPC-C and TPC-H type workloads, and it can be used with SQL Server, Oracle, DB2, MySQL, PostgreSQL, Redis and Trafodian. For Microsoft Exchange we have the Jetstress tool which can be used for similar purposes.
Your local Solutions Architects stand ready to take any query you may have, pre and post testing.
Download the MaxParallel for Windows Server Free Trial in your own environment and get performance improvements in minutes!