Which HDD for Content Applications—How and What to Test: Part 2
This is the second in a multi-part series based on a white-paper hands-on lab report I did compliments of Servers Direct and Seagate.
Join the DZone community and get the full member experience.Join For Free
The focus is looking at the Servers Direct (www.serversdirect.com) converged Content Solution platforms with Seagate Enterprise Hard Disk Drive (HDD‘s). In this post, we look at some decisions and configuration choices to make for testing content applications servers as well as project planning.
Content Solution Test Objectives
In a short period, collect performance and other server storage I/O decision-making information on various HDD‘s running different content workloads.
Working with the Servers Direct staff, a suitable content solution platform test configuration was created. In addition to providing two Intel-based content servers, Servers Direct worked with their partner Seagate to arrange for various enterprise-class HDD’s to be evaluated. For these series of content application tests, being short on time, I chose to run some simple workloads including database, basic file (large and small) processing, and general performance characterization.
Content Solution Decision Making
Knowing how Non-Volatile Memory (NVM) NAND flash SSD (1) devices (drives and PCIe cards) perform, what would be the best HDD-based storage option for my given set of applications? Different applications have various performance, capacity, and budget considerations. Different types of Seagate Enterprise class 2.5” Small Form Factor (SFF) HDDs were tested.
While revolutions per minute (RPM) still play a role in HDD performance, there are other factors including internal processing capabilities, software or firmware algorithm optimization, and caching. Most HDDs today have some amount of DRAM for read-caching and other operations. Seagate Enterprise Performance HDDs with the enhanced caching feature (2) are examples of devices that accelerate storage I/O speed vs. traditional 10K and 15K RPM drives.
Project Planning and Preparation
Workload to be tested included:
- Database read/writes
- Large file processing
- Small file processing
- General I/O profile
Project testing consisted of five phases, some of which overlapped with others:
Phase 1 – Plan
Identify candidate workloads that could be run in the given amount of time; determine time schedules and resource availability; create a project plan.
Phase 2 – Define
Hardware define and software define the test platform.
Phase 3 – Setup
The objective was to assess plug-play capability of the server, storage, and I/O networking hardware with a Linux OS before moving on to the reported workloads in the next phase: initial setup and configuration of hardware and software; installation of additional devices along with software configuration; troubleshooting and learning as applicable. This phase consisted of using Ubuntu Linux 14.04 server as the operating system (OS) along with MySQL 5.6 as a database server during initial hands-on experience.
Phase 4 – Execute
This consisted of using Windows 2012 R2 server as the OS along with Microsoft SQL Server on the system under test (SUT) to support various workloads. Results of this phase are reported below.
Phase 5 – Analyze
Results from the workloads run in phase 3 were analyzed and summarized in this document.
(Note 1) Refer to Seagate 1200 12 Gbps Enterprise SAS SSD StorageIO lab review
Planning and Preparing the Tests
As with most any project, there were constraints to contend with and work around.
Test constraints included:
- Short time window
- Hardware availability
- Amount of hardware
- Software availability
The three most important constraints and considerations for this project were:
- Time – This was a project with a very short time “runway,” something common in most customer environments who are looking to make knowledgeable server/storage I/O decisions.
- Amount of hardware – Limited amount of DRAM main memory, sixteen 2.5” internal hot-swap storage slots for HDDs as well as SSDs. Note that for a production content solution platform, additional DRAM can easily be added, along with extra external storage enclosures to scale memory and storage capacity to fit your needs.
- Software availability – Utilize common software and management tools publicly available so anybody could leverage those in their own environment and tests.
The following content application workloads were profiled:
- Database reads/writes – Updates, inserts, read queries for a content environment
- Large file processing – Streaming of large video, images, or other content objects.
- Small file processing – Processing of many small files found in some content applications
- General I/O profile – IOP, bandwidth, and response time relevant to content applications
What This All Means
There are many different types of content applications ranging from little data databases to big data analytics as well as very big fast data such as for video. Likewise, there are various workloads and characteristics to test. The best tests and metrics are those that apply to your environment and application needs.
Continue reading part three of this multi-part series here looking at how the systems and HDDs were configured and tested.
Published at DZone with permission of Greg Schulz, DZone MVB. See the original article here.
Opinions expressed by DZone contributors are their own.