Customer want to deployment Solr Cloud on AWS EC2 ask AWS to provide benchmark data about EBS volume which target IOPS required > 30,000 and Solr engine need SSD volumes.
Benchmark EBS Volumes official guide
- Launch an EBS-optimized instance.
- Create new EBS volumes.
- Attach the volumes to your EBS-optimized instance.
- Configure and mount the block device.
- Install a tool to benchmark I/O performance.
- Benchmark the I/O performance of your volumes.
- Watch the cloudwatch metrics
- Delete your volumes and terminate your instance.
For example, two 500 GiB Amazon EBS io1 volumes with 4,000 provisioned IOPS each will create a 1000 GiB RAID 0 array with an available bandwidth of 8,000 IOPS and 1,000 MB/s of throughput.
Creating a RAID 0 array allows you to achieve a higher level of performance for a file system than you can provision on a single Amazon EBS volume. A RAID 1 array offers a "mirror" of your data for extra redundancy.
RAID 5 and RAID 6 are not recommended for Amazon EBS because the parity write operations of these RAID modes consume some of the IOPS available to your volumes.
You can creat a RAID array with hight volumes such as (we recommend 6 volumes), which offers a high level of performance. But for this customer case testing, 4 volumes RAID will used.
注意
- Raid0和Instance store对于数据持久性都有局限，需要通过定期备份到S3。考虑到Solr Cloud分布式特性以及对于IOPS的要求，因此可以使用。
- 如果由于实例的IOPS和吞吐有上限，因此需要更大IOPS和吞吐，需要开更大的机型，相应的机器性能更好。
- 如果想不改变单盘的IOPS提升总体IOPS，这个需要做raid，但是带来的问题也是
Performance of the stripe is limited to the worst performing volume in the set以及Loss of a single volume results in a complete data loss for the array。 比如要40000的IOPS，可以raid 4块 10000IOPS的盘；也可以不做raid，然后直接1块或者多块 40000IOPS的盘 - GP2是突增类型，注意credit。
- Raid0和Raid1的选择 The following table compares the common RAID 0 and RAID 1 options.
| Configuration | Use Advantages | Disadvantages |
|---|---|---|
| RAID 0 | When I/O performance is more important than fault tolerance; for example data replication is already set up separately | I/O is distributed across the volumes in a stripe. If you add a volume, you get the straight addition of throughput and IOPS |
| RAID 1 | When fault tolerance is more important than I/O performance; for example, as in a critical application. | Safer from the standpoint of data durability. |
- Raid0 on instance storage for different instance type
- i3.4xlarge 2 x 1900GiB NVMe SSD Raid0 on instance storage
- m5d.8xlarge 2 x 600GiB NVMe SSD Raid0 on instance storage
- Compare io1 high IOPS v.s gp2 RAID0 v.s io1 RAID0 Note: If you need IOPS and high throughput you need use the large instance type to avoid instance limits
- m5.8xlarge io1 2048 provision IOPS 40000
- m5.12xlarge gp2 3000GiB x 4 Raid0 provision IOPS 36000
- m5.12xlarge io1 2048GiB provision IOPS 40000
- m5.12xlarge io1 500GiB x4 Raid0, provision IOPS 44000
- Testing snapshot
| instance type | IOPS randread 16 KB I/O | Throughput (MBps) randread 16 KB I/O | IOPS randwrite 16 KB I/O | Throughput (MBps) randwrite 16 KB I/O | IOPS randread 4 KB I/O | Throughput (MBps) randread 4 KB I/O | IOPS randwrite 4 KB I/O | Throughput (MBps) randwrite 4 KB I/O |
|---|---|---|---|---|---|---|---|---|
| m5.8xlarge | 30327 | 485 | 30139 | 482 | N/A | N/A | N/A | N/A |
| m5.12xlarge io1 | 40308 | 644.9 | 40216 | 634.4 | N/A | N/A | N/A | N/A |
| m5.12xlarge gp2 raid0 | 36299 | 580.8 | 36155 | 578.5 | N/A | N/A | N/A | N/A |
| m5.12xlarge io1 raid0 | 40295 | 644.7 | 40219 | 643.5 | N/A | N/A | N/A | N/A |
| m5d.8xlarge Single NVMe SSD | 205133 | 3205.3 | 32233 | 515.7 | 437578 | 1709.3 | 117463 | 469.8 |
| m5d.8xlarge NVMe SSD Raid0 | 135220 | 2112.9 | 64470 | 1007.4 | 470710 | 1838.8 | 234654 | 938.6 |
| I3.4xlarge Single NVMe SSD | 216421 | 3381.7 | 48958 | 783.3 | 581387 | 2271.5 | 180559 | 722.2 |
| I3.4xlarge NVMe SSD Raid0 | 243184 | 3799.8 | 97678 | 1526.3 | 824841 | 3222.4 | 357232 | 1395.5 |
optimize the performance of my Amazon EBS Provisioned IOPS volumes
AWS IO Performance: What’s Bottlenecking Me Now? It not just the volume