Amazon Web Services EBS Volumes provides the block level storage volumes of Amazon EC2 instances. EBS volumes are highly available and reliable storage volumes that can be attached to the EC2 instances of the same Availability Zone over AWS internal network. Elastic Block Storage (EBS) is a storage service that is backed by network-connected block storage. So IOPS performance of an EBS volume is measured by the network throughput between the instance and EBS volume.
Two kinds of EBS volumes available
Standard Volumes offer cost-effective storage and can deliver approximately 100 IOPS on average, or burst of hundreds of IOPS. Standard Volumes size can range between 1 GB to 1 TB.
Provisioned IOPS (PIOPS) Volumes are designed for heavy I/O intensive workloads like databases and consistent high I/O throughput. You can specify an IOPS rate when you create the volume. PIOPS Volumes size can range between 10 GB and 1 TB.
The Standard or Provisioned IOPS Volumes I/O performance will depend on the Volume Size and Instance Type.
RAID on EBS Volumes
To get both a faster I/O performance and more than 1TB of volume sizes, you can create a RAID 0 striping using more than one EBS volume.
Advantage: When we perform the RAID 0 Striping of multiple volumes, IOPS are distributed among the volumes of a stripe. If you add another volume to RAID 0, you get the straight addition of IOPS throughput of that volume and additional volume size.
To take advantage of the RAID 0 striping, the instance type should support the High Networking Performance (like m3.xlarge, c3.2xlarge, r3.2xlarge, etc.).
Disadvantage: Loss of a single volume results in a complete data loss.
To avoid the data loss, we should take Snapshots of the RAID volumes frequently, and suspend the writes while taking the snapshot to ensure data integrity and reliability.
Here we will discuss on how to configure the RAID 0 Striping on 2 EBS volumes of several Linux machines. You can expand this to any number of volumes as per your requirements.
Create RAID Configuration
- Create the two EBS volumes with a size of 5 GB and attach them to the Amazon Linux Instance.
- Create a RAID 0 striped array using the two attached volumes:
mdadm –create /dev/md0 –level=0 –level=stripe –raid-devices=2 /dev/xvdc /dev/xvdd
(Assuming the 2 EBS volumes are attached as /dev/xvdc and /dev/xvdd respectively)
- Format the raid volume with your favorite file system:
mkfs.ext4 /dev/md0 - Mount it on a temporary directory:
mount /dev/md0 /srv/raid0 - Create some dummy data on /srv/raid0 directory:
dd if=/dev/zero of=file1 bs=1024 count=65530 (run it about 10 times)
It will fill 9GB of information onto the RAID meta device. Both the volumes contains data, given that each volume size is 5 GB.
Snapshot RAID volumes
- Create snapshots of the two volumes after suspending writes on those volumes.
Restore RAID Volumes
- Create the new EBS volumes from the snapshots we have created before.
- Launch an EC2 instance of either Amazon Linux or CentOS and attach the newly created EBS volumes.
- You can see the RAID meta device is created automatically by the OS itself. Use the fdisk -l command to check it out. Can you mount on a temporary mount point and verify whether 9GB of data is available or not?
- Now, launch an EC2 instance of Ubuntu and attach the newly created EBS volumes.
- Run the fdisk –l command and verify whether the RAID meta device created automatically by the OS itself. If not, run the following command to create the RAID meta device
madam –assemble /dev/md0 /dev/xvdc /dev/xvdd - Now you can see a meta device with the name /dev/md0. You can mount it and check its content.
Hope this article will help you to manage the RAID configuration on EBS volumes to gain the maximum IOPS.
