Working with the suite of AWS services requires making a lot of choices. While having options is a good thing, you need to fully understand the services available to you in order to make decisions about your business infrastructure. Incorrect choices can result in a suboptimal environment which will affect the efficiency of your workloads and incur unnecessary costs.
Choices are commonly made around S3 storage classes or EC2 instance types. Making EBS volume choices is not as common, since most developers, by default, simply go for the standard, general purpose EBS volume types when provisioning an instance. Nevertheless, there are other options that can and should be considered. This blog post will look at the most recent volume offerings to help you by doing an EBS volume comparison to select the best one for your needs.
What Is EBS?
Amazon Elastic Block Store (EBS) is a high-performance block storage solution designed to be used with EC2 instances. It works by giving the user access to a multitude of small network-attached blocks, each of which is presented as a simple volume. This distributed way of providing storage allows for easy expansion of storage capacity when needed; EBS volumes can scale into petabytes of data. EBS also ensures high availability (HA) and durability by replicating all data within the Availability Zone (AZ).
EBS volumes are very easy to use, whether you need to add them to or remove them from the EC2 instance, modify their size, or change their type. With only a couple of clicks, you can create an EBS snapshot, effectively making a backup of your entire volume. This can later be used to recreate the data if necessary.
The block storage offered by EBS volumes can cover any workload, from the most demanding mission-critical applications requiring high throughput to large sequential jobs like data warehousing and log processing. If the provided performance is not enough, you can also combine multiple volumes on a single instance to achieve even better results. EBS provides a very secure data storage solution, since it was built with compliance in mind. Each volume allows for in transit, at rest, and backup encryption. EBS volumes are also very cost-effective.
When choosing your EBS volume types, you’ll find multiple options. The remainder of this post is devoted to examining them.
An Overview of EBS Volume Types
All EBS volumes fall into two categories: SSD-backed volumes and HDD-backed volumes. SDD-backed volumes are designed for transactional workloads where IOPS (input/output operations per second) is the most important consideration. HDD-backed volumes are meant for streaming workloads, where throughput is more important than IOPS. EBS volumes also come in various different types, each designed for a different use case and set of requirements.
General Purpose SSD (gp2) Volumes
General Purpose SSD (gp2) volumes are the most common volume type. They were designed to be a cost-effective storage option for a wide variety of workloads. Gp2 volumes cover system volumes, dev and test environments, and various low-latency apps. They come in sizes between 1Gib and 16Gib and provide very low latency, down to single-digit milliseconds. They have a decent IOPS (starting from 100 and going all the way to 16000 IOPS) and a maximum throughput of 250MiB/s. You can combine multiple EBS volume types in a RAID to achieve even higher performance on a single instance.
The overall performance of gp2 volumes is directly related to their size; users get 3 IOPS per each GiB. This defines baseline performance as well as how quickly you gain the I/O credits which can be used for burst performance when needed.
Gp2 volumes are fairly cheap, especially for the balanced performance they provide. They are priced at $0.1 per GB per month of provisioned storage.
Provisioned IOPS SSD (io1) Volumes
Provisioned IOPS SSD (io1) EBS volume types are a special type of volume created to fulfill the needs of very intensive I/O workloads that require very high throughput. They are useful for cases which are latency-sensitive, like large database workloads (e.g., MySQL, Cassandra, MongoDB, and Oracle) and critical business applications that need the kind of sustained performance gp2 volumes can’t achieve.
Io1 volumes can store between 4GiB and 16TiB, and their IOPS can be as low as 100 or high as 64000 IOPS per volume with up to 1,000 MiB/s of throughput. While the performance of gp2 volumes is dictated by volume size, the performance of provisioned IOPS SSD volumes can be set during creation time. It is limited by a maximum IOPS to volume size ratio of 50:1.
Of course, all of this comes with a price—$0.125 per GB per month of provisioned storage and $0.065 per provisioned IOPS per month. This can amount to a costly solution in some cases, so be careful when provisioning these volumes.
Throughput Optimized HDD (st1) Volumes
Throughput Optimized HDD (st1) volumes are a type of volume that offers low-cost storage while fulfilling the need for sequential workloads that require more throughput than IOPS. When working with data warehouses, log processing, ETL (extract, transform, load) or AWS EMR, this is a volume type to look into. Keep in mind that this volume type cannot be used as a boot volume.
St1, like gp2, relies on burstable performance, and volume size will be the main factor when calculating baseline performance. St1 volumes can range in size between 500GiB and 16TiB, and they allow for 500 MiB/s of throughput per volume with 500 IOPS.
St1 volumes also come with a lower price tag: $0.045 per GB per month of provisioned storage.
Cold HDD (sc1) Volumes
Cold HDD (sc1) volumes, like st1 volumes, provide low-cost storage for workloads that rely on throughput rather than IOPS. Sc1 volumes are primarily used for large amounts of data that is infrequently accessed, or in cases where the cost of storage is the most important factor. Sc1, just like st1, cannot be used as a boot volume and relies on burstable performance.Sc1 volumes come in sizes between 500GiB and 16TiB and provide up to 250 IOPS and 250MiB/s of throughput per volume.
Sc1 volumes are the cheapest option available, costing only $0.025 per GB per month of provisioned storage.
EBS Volume Types Table
The table below provides a quick comparison of the EBS pricing and volume types:
|
Volume Type |
General Purpose SSD (gp2) |
Provisioned IOPS SSD (io1) |
Throughput Optimized HDD (st1) |
Cold HDD (sc1) |
|
Uses |
General purpose SSD volume that balances performance and price; used for low-latency app, dev and test environments |
Highest-performance SSD volume, designed for mission-critical, low-latency, or high-throughput applications; used for large databases, and critical business applications |
Low-cost HDD volume designed for throughput-intensive workloads; used for log processing, data warehouses, and streaming workloads |
Lowest cost HDD volume designed for infrequently accessed data; used as a cheap storage solution |
|
Volume Size |
1GiB—16TiB |
4GiB—16TiB |
500GiB—16TiB |
500GiB—16TiB |
|
Max Throughput per Volume |
250MiB/s |
1000MiB/s |
500MiB/s |
250MiB/s |
|
Max IOPS per Volume |
16000 |
64000 |
500 |
250 |
|
Cost |
$0.1 per GB |
$0.125 per GB + $0.065 per IOPS |
$0.045 per GB |
$0.025 per GB |
Instance store volumes
Instance store volumes are a special block-level storage option that can be used on AWS. Unlike EBS volumes, which are network attached, these are physically attached to the host computer. They are meant to serve as temporary storage solutions only, used for caches, buffers, and other ephemeral data.
The lifetime of the instance store volume depends directly on the instance to which it is attached. If you reboot the instance, the data will persist; but, if you stop the instances, terminate them, or even change the instance type, your data will be lost.
Instance store volumes can actually provide very high performance, delivering high random I/O which can be good for low-latency workloads. They can even outperform EBS volumes in writes and uncached reads.
Instance store volumes come with no individual cost; instead, they are included as part of the instance’s usage cost. They also come with specific instance types, usually specialized ones with higher price tags. For example, the i3 family (starting at $0.156 per hour), a storage-optimized instance designed for I/O performance, uses instance store volumes for its workloads. High-performance, GPU-backed instances like g3 ($1.14 per hour) and memory-optimized instances meant for in-memory applications such as x1 (a very expensive instance with its lowest possible cost at $6.669 per hour) rely on instance store volumes. They all benefit from the great performance that these ephemeral, locally-attached volumes provide, and, because they were created with specific use cases in mind, they have very few drawbacks.
EBS Volumes: Doing your Research and Choosing the Best Volume
Over time, the block storage offerings available on AWS have grown substantially. Users can now choose from a variety of options based on performance or cost priorities. For standard workloads, gp2 volumes offer balanced performance while keeping costs down. For sequential workloads, st1 volumes can perform very well while reducing the cost even further. Infrequently accessed data can be stored on sc1 volumes—the cheapest option you can get. And, of course, io1 volumes are there for the workloads that require the highest possible performance.
This variety of options has the potential to greatly benefit your business environment, so doing your research and choosing the best volume for your environment is well worth your time.
Click here to read original post ->
