Delivering higher levels of performance from cloud-based storage usually drives up the performance of the entire application stack, as business critical data in enterprise organizations must always be read and written to at scale. Native cloud storage, such as AWS EBS, is the logical place to start when you need to add block-level storage to cloud servers, however, can direct access to this storage actually decrease the maximum potential performance available to us?
Cloud Volumes ONTAP is the data management platform from NetApp that builds on top of Amazon EBS, delivering higher levels of storage performance through a sophisticated set of performance enhancing features. Using Cloud Volumes ONTAP, you can boost disk performance for both random and sequential I/O, while at the same time gaining access to a wide range of data management capabilities, such as multi-region data protection and storage efficiencies that decrease your cloud storage operational expenditure.
In this article we will look in detail at the performance features of Cloud Volumes ONTAP, and demonstrate how they can improve I/O performance compared to raw Amazon EBS disks by reviewing the results of a performance benchmark test conducted by NetApp.
Amazon EBS is a service that provides block-level storage within the Amazon cloud, allowing users to allocate that storage using a variety of disk types, depending on whether they require performance or capacity. The storage provided by Amazon EBS is redundant within an Availability Zone, and AWS snapshots can be used to perform backups to Amazon S3.
Cloud Volumes ONTAP is a multicloud data platform that brings NetApp’s extensive background in enterprise storage to AWS, Azure, and GCP. By using the native cloud resources available in each environment, storage architects can tailor the deployment of Cloud Volumes ONTAP to fit their precise requirements.
For example, when deploying a new single node or high availability (HA) deployment of Cloud Volumes ONTAP, you can choose whether to use general purpose (m4/m5), compute optimized (c4/c5), or memory optimized (r4/r5) Amazon EC2 servers. The underlying disk storage can be configured to use all of the available AWS storage types, with the added ability to non-disruptively move data between them, as well as transparently tier capacity storage over to Amazon S3 on a block level. For Amazon EBS storage, users have control over the type and size of the disks they use, with the option of pooling disks together to create RAID groups, or to create different tiers of storage pools.
In addition to options available for configuring the cloud infrastructure you wish to use, Cloud Volumes ONTAP offers a number of performance-enhancing features that increase all levels of Amazon EBS performance that you can achieve.
Cloud Volumes ONTAP’s FlashCache features allow recently used data to be intelligently cached on local attached NVMe storage. Thanks to caching IOs on NVMe, it is possible for read requests to be served without hitting the underlying disks (EBS in this case). This dramatically improves the speed of random reads, such as those performed by databases, email servers, and file services. The maintenance of the cache is performed in real time by Cloud Volumes ONTAP, and so the use of this feature is transparent to end users. This gives users NVMe performance with gp2 costs.
Cloud Volumes ONTAP allows for the configuration of multiple data LIFs (Logical Interfaces) for communicating with clients, thereby enabling the use of multi-path I/O for maximizing data throughput. This can have a very positive impact on large volume sequential reads, as all paths the back-end storage can be used in tandem, thereby increasing the available bandwidth for accessing data.
In scenarios where there is an overarching need to increase sequential write performance, Cloud Volumes ONTAP provides the High Write Speed setting, which can return success to an application writing data directly from the cache, before having actually committed that data to stable storage. This leaves open a small window for data loss, as in a case where a failure occurs just before the write to stable storage is completed.
This high speed write feature is intended for use in domains where some potential data loss is acceptable, such as data streaming applications, or where the data would be recoverable at the application layer, such as in a multi-node clustered architecture.
The following section provides details of a performance comparison of Cloud Volumes ONTAP and Amazon EBS carried out by NetApp. The test environment consisted of a client host accessing 4 TB of General Purpose SSD (gp2) storage from Cloud Volumes ONTAP, as compared to accessing that storage directly from Amazon EBS.
AWS EBS vs Cloud Volumes ONTAP: Test Environment Configuration.
Looking at the results for random read activity, we see a massive performance improvement made to latency, throughput, and Amazon EBS IOPS through the use of Cloud Volumes ONTAP FlashCache. There is an almost eightfold increase in read IOPS when the FlashCache is being fully leveraged:
Using the Cloud Volumes ONTAP High Write Speed setting, we see another significant increase in performance, in this case for sequential write activity. The Cloud Volumes ONTAP FlashCache can be configured to ensure that write operations do not empty the cache and reduce its effectiveness, but only allow sequential read and writes to warm up the cache.
Cloud Volumes ONTAP delivers significant improvements to Amazon EBS performance as a result of powerful features, such as FlashCache. As shown in the performance comparison results, FlashCache provides a manifold increase in general purpose SSD IOPS, as well as reduced latency and increased throughput.
Above and beyond I/O performance, Cloud Volumes ONTAP also gives storage architects an entire suite of data management capabilities, including data protection, file sharing services, and storage efficiency technologies that help reduce your spend on cloud storage.