The Infinidat InfiniBox series are high performing storage systems and deliver excellent performance. However, every controller has limitations, so it’s good to understand the key metrics that describe the health of the system. This blog covers the key metrics driving Infinidat InfiniBox system performance and health.
Key Workload Factors for Infinidat InfiniBox
At the surface level, Infinidat InfiniBox’s key metrics are similar to most storage systems, but understanding how Infinidat’s nuances separate it from other storage systems in your multi-vendor open-system helps when configuring your environment optimally for Infinidat systems.
The key workload factors driving Infinidat InfiniBox performance are the I/O Rates, Read to Write Ratio, Fast Writes, and Transfer Sizes.
I/O’s Per Second
The I/Os per second is the number of Input / Output requests that the system can service. The ability of the system to service I/Os is dependent on a number of workload factors such as: read to write ratio, size of I/O, sequential vs random nature of the workload and the locality of reference of the data which contributes to the cache read hit ratio.
Read to Write Ratio
The read to write ratio is important because reads require less work for the storage system as they are often satisfied in one of the cache layers. For reads, the Infinidat InfiniBox® examines the metadata to determine if the data resides in global DRAM, SSD cache or on SATA drives. Each layer is progressively slower.
Infinidat has a massive global DRAM and uses intelligent algorithms to satisfy the vast majority of read requests from DRAM or SSD cache. The DRAM cache hits and SSD cache hits are important metrics in understanding how the workload is performing.
Writes are written to a single large pool of RAM based cache that is shared across the three node Active-Active-Active architecture. As the written blocks “cool down” Infinidat InfiniBox® cherry picks them for destage from the RAM cache in order to write them efficiently to disk.
One of the things that Infinidat InfiniBox® achieves with their numerous write optimization techniques is that they are able to achieve a high percentage of write hits, that is, updating writes in cache and cutting down on the number of destages to disk from cache.
Data blocks are stored on persistent media in sections of 64KB for user data and are written using stripes of 14 data blocks and 2 parity sections (14+2). A total InfinRAID stripe is 68KB X 16 or 1088 KB stripe across 16 independent drives. Leveraging various write optimization techniques, the Infinidat InfiniBox® attempts to destage data sequentially, even if it arrived randomly, significantly improving throughput for write operations.
Even so, there are limitations on every storage system, and it is important to monitor your effective read/write ratios and overall write throughput.
Workloads with a locality of reference that can be contained within the capacity of the fastest layer will provide the best performance while workloads with a locality of reference much larger than the DRAM, or SSD cache will reside on the SATA drives and will take longer to fulfill.
Infinidat InfiniBox Latency / Response Time
Latency, or response time, is the amount of time it takes for the system to service an I/O request. As described previously, there are several layers within the system, and they all provide low latency as compared with SATA drives.
The following table provides the latency and compares it to the relative effective distance from the processor at the speed of light. The further from the processor, the longer the I/O request takes.
|Technology||Latency||Effective Distance from the Processor at the Speed of Light|
|DRAM||1 ns||12 inches|
|NVMe SSD||7-150 us||1.32 to 28.4 miles|
|SATA Drives||10 ms||1894 miles|
Information derived in connection with: https://formulusblack.com/blog/compute-performance-distance-of-data-as-a-measure-of-latency/
How to Optimize Infinidat InfiniBox Performance with End-to-End Visibility
Even with a comprehensive understanding of Infinidat’s key performance metrics, effectively monitoring and managing your InfiniBox® storage arrays alongside the rest of your multi-vendor environment requires deeper visibility across your whole system from your VMware through your fabric to your Infinidat InfiniBox.
This includes interpretation of all the key performance metrics:
- Throughput (factor of I/O rate and Transfer size)
- Front-end Read Response Time
- Front-end Write Response Time
- Front-end Overall Response Time
- Read Hit %
- Port Read Response Time
- Port Write Response Time
IntelliMagic Vision is a multi-vendor SAN infrastructure performance and health monitoring solution that utilizes built-in deep expert knowledge about each of the supported platforms, such as Infinidat InfiniBox, and artificial-intelligence derived insights to provide a single pane of glass for optimizing and protecting the performance and availability of your entire distributed-system environment.
The image above shows one of the out-of-the-box interactive reports presenting all of your Infinidat InfiniBox® key metrics in a single view. By clicking on any of the individual charts in the multi-chart, you drill down to a specific port or volume for quick and seamless root cause analysis.
End-to-End FlashArray Connectivity and Configuration Analysis
IntelliMagic Vision also provides end-to-end visibility for performance and configuration analysis from your hosts through your fabric to your Infinidat InfiniBox® so that you can quickly identify if there are any connectivity configuration issues such as the single path from the host to the storage array as depicted in the image below.
Efficient Capacity Planning for Infinidat Systems
The Infinidat InfiniBox® has a number of technologies deployed to reduce the footprint and efficiently use the capacity on the system. IntelliMagic Vision provides insight into how efficiently capacity is being used.
In the following table, IntelliMagic Vision provides the utilized capacity along with the data reduction ratio which is key in understanding how efficiently the system is reducing your data.
In addition to showing the key capacity metrics, IntelliMagic Vision helps in planning for future storage requirements through its capacity forecasting views. In the chart below the ‘Time until running out’ is shown in red. When it is above 360 days it is not displayed on the chart.
Better Infinidat Configuration and Monitoring
Infinidat InfiniBox, just with any platform, has key nuances related to its key metrics that are crucial to understand to ensure optimal performance and availability. It is worth taking the time to understand how the nuances of Infinidat storage arrays impact your environment as a whole.
IntelliMagic Vision offers complete end-to-end performance, configuration and capacity monitoring and visibility for the Infinidat InfiniBox® as well as all of the top storage systems on the market. We apply artificial intelligence to the hardware-provided optimal configuration requirements with your specific workload needs to help you proactively eliminate availability and risks and ensure your environment is optimally configured.
If you are looking for in-depth support of your Infinidat InfiniBox® or need a better way to manage your multi-vendor storage environment, start a free trial of IntelliMagic Vision today.
Subscribe to our Blogs
Subscribe to our newsletter and receive monthly updates about the latest industry news and high quality content, like webinars, blogs, white papers, and more.
Understanding Pure Storage FlashArray Key Metrics
Performance bottlenecks on your storage array can impact your applications ability to service requests in a timely fashion. It’s important to have the right visibility into your storage systems to proactively identify and resolve bottlenecks.
Improving IT and Business Organization Alignment through Dynamic Collaboration
Silos have the benefit of grouping logical functional areas within the larger organization in order to provide deep and comprehensive support. Learn 4 ways to move your teams from unilateral to collaborative communication.
Avoid Islands In Your Data Lakes
Executives have been asking for a holistic view of the enterprise applications and infrastructure for years. The result has been a slew of overly complex, hard to maintain, expensive, and siloed IT management solutions.