We did a similar comparison between the x1.16xlarge and r3.8xlarge in a previous article, but let’s see what additional insights come from comparing the x1.16xlarge with the r4.16xlarge. Let’s walk through their specs and pricing to highlight where your compute spending could be going from an AWS cost efficiency perspective.
Specs and pricing
The r4.16xlarge provides dual-socket, high-frequency Intel Xeon E5-2686 v4 (Broadwell) processors DDR4 memory and has support for AWS’s Enhanced Networking up to 20 Gbps. The R4s don’t feature SSDs like the previous generation of R3 provide. Instead, you must attach EBS volumes to each instance. Attached EBS volumes can take advantage of network optimizations delivering up to 12 Gbps of dedicated throughput.
Each x1.16xlarge instance features two Intel Xeon E7 8880 v3 (Haswell) processors running at 2.3GHz. Each instance provides 32 cores and 64 vCPUs. There’s 976 GB of RAM (featuring Single Device Data Correction, or SDDC+1) onboard, with 1,920 GB of SSD storage and 10 Gbps of network bandwidth throughput available. The x1.16large features optimized bandwidth to EBS volumes (up to 5 Gbps) at no additional cost.
|Instance||Compute Units (ECU)||Memory (GB)||Storage (GB)||Network performance||EBS optimization|
|r4.16xlarge||195 units||488 GB||EBS only||20 Gbps||12,000 Mbps|
|x1.16xlarge||174.5 units||976 GB||1,920 GB (SSD)||10 Gbps||5,000 Mbps|
Pricing and cost efficiency
|Instance||Linux On-Demand Pricing||Cost per ECU per hour||Cost per GB memory per hour|
|r4.16xlarge||$4.256 per hour||$0.0218 per ECU per hour||$0.0087 per GB per hour|
|x1.16xlarge||$6.669 per hour||$0.0382 per ECU per hour||$0.00683 per GB per hour|
The r4.16xlarge provides a lower amount of memory, but delivers more cost-efficient computing
Not only does the x1.16xlarge provide 50% more memory, but you can access it all at a 22% cheaper rate than the r4.16xlarge. However, the r4.16xlarge features a higher amount of available computing units (ECU) using a newer generation of Intel processor. It provides these compute units at a cheaper rate per ECU, being 43% cheaper than the x1.16xlarge.
The r4.16xlarge won’t have persistent data without an attached EBS volume
The x1.16xlarge provides 1,920 GB of onboard SSD storage, but this could be an issue if you constantly cycle these instances on and off via EC2 usage optimization strategies or autoscaling. The data will not persist when the instance gets shut off. However, writing data to an attached EBS volume is a solution to this, and the x1.16xlarge features EBS optimization up to 5,000 Mbps.
Any r4.16xlarge instance will require attaching an EBS volume as storage. While this adds an additional cost to instance usage, you can take advantage of the r4.16xlarge’s 12,000 Mbps of EBS optimization. Keep in mind that If you need more I/O speed, provisioning IOPS will add a bit more to your EBS billing.
Which is right for which job?
Both instance families are great candidates for running memory-intensive workloads. But there are some nuances to what each family is optimized for, according to AWS.
- The X1 family is ideal for running in-memory databases (e.g. SAP HANA), big data processing (Apache Spark, Presto), and works well with high-performance computing apps.
- The R4 family specializes in workloads that involve scaling in-memory caching, high-performance databases, high-performance business intelligence tasks, high-performance data ingestion, data mining, big data analytics, Hadoop or Spark clusters and working with the analysis of unstructured collections of data.
With this in mind, in addition to knowing which instances provide cost efficiencies between computing power and memory, you have a better idea of what this kind of spending on AWS can lead to.
Monitor your EC2 cost and usage data to make the best choice
Using a cloud cost management tool, like Cloudability, can help you take a closer look at how much waste or value there is within your AWS infrastructure. In the case of EC2, you can monitor whether you’re using an x1.16xlarge or an r4.16xlarge and determine if you’re putting all those compute units and GBs of RAM to work.
Once you make adjustments to be as efficient as possible with EC2, purchasing Reserved Instances for those hours can lower your bill significantly. When working with relatively expensive instances like the x1.16xlarges and r4.16xlarges, these savings can add up.
To see this type of cloud cost management and EC2 optimization at work, please get in touch and we can set you up with a free trial today, or ask our EC2 experts about how to choose the right instances and save for the long-haul with your current infrastructure.