Pure Storage’s FlashRecover introduces tremendous value to enterprises in three key areas: Performance, simplicity, and scalability.
There’s value in finishing backups faster. Backups offer protection services to production applications and data. With data and application sets growing rapidly, creating backups in a timely fashion has become more difficult. By being able to back up more often and write quicker, a business can meet growing demand while meeting the most stringent recovery point objectives (RPOs). While an “incremental forever” approach to backups can shrink backup windows, during recoveries there are no such optimizations. Restore performance is often dictated by how fast you can read from your backup target and how efficiently your backup target file system is designed. Flash eliminates defragmentation-related concerns and inherently brings performance to restores. A high-throughput
FlashBlade system offers massive recovery capabilities, allowing the recovery of databases or petabyte-scale environments with much more rapid speed compared to a traditional backup application while keeping the whole backup ecosystem operational very simple.
FlashRecover Oracle Protection Solutions Architecture Overview
The source application Oracle databases can be hosted on any storage, but for fast backup and fast restore, it is imperative that it is hosted on an all-flash array like Pure Storage FlashArray™. For this test, four Oracle database servers are hosted across three FlashArray systems (with 1-1-2 databases per FlashArray system), that are connected to four ESX hosts over iSCSI. Datastores are created on the three FlashArray systems, and virtual machine disks (VMDKs) are created for every Oracle database server to host the database. This blog will only illustrate the performance measurement using four Oracle databases. Datastores are created from the three FlashArray systems, and the VMDKs are created for every Oracle database server to host the database. Pure FlashRecover, Powered by Cohesity architecture is shown in the following figure.
Every ESX server is equipped with a 10Gbps network interface, which enables approximately 1 GBps of bandwidth between the database servers, FlashRecover cluster, and FlashBlade. Hence, running a single database backup/restore on this environment can only yield a maximum of 1 GBps of bandwidth. To demonstrate higher and scalable bandwidth, multiple database backups and restores across ESX servers are performed.
Series of full backups of the Oracle databases are performed through FlashRecover UI by varying the number of databases from 1, 2, and 4 with 16 channels. To simulate the real-world scenario, after the first level zero backup, a 5% of change is introduced in the database prior to the next level one full backup. The backup speed is captured from Oracle’s internal view and hence this shows the end-to-end backup timing.
Based on our environment set up where each database host is limited to 10 Gbps, single database backup achieved on an average of 811 MBps with 16 channels.
A series of full restores of the Oracle databases are performed through FlashRecover UI by varying the number of databases from 1, 2, and 4. The restore speed is captured from Oracle’s internal view, and, hence, this shows the end-to-end restore timing.
For more information on this solution and if interested to read the complete solutions whitepaper please visit this link FlashRecover Solutions.
To view other blogs on Flashrecover click here.
