When designing high-performance workloads on AWS Outposts racks, understanding the interplay between compute instances and storage performance is critical to avoid bottlenecks and optimize costs. AWS Outposts extends AWS infrastructure, services, and tools to on-premises environments, enabling ultra-low latency, local data processing, and data residency compliance. However, final IOPS and throughput depend on the correct selection of EC2 instance type, EBS volume type, and multi-volume configuration. At Q2BSTUDIO, a software development and technology company specializing in cloud architectures, we help clients design solutions that maximize Outposts performance, integrating cloud AWS/Azure services, artificial intelligence, cybersecurity, and BI/Power BI.
The first step in any design is understanding the selected EC2 instance's capability. AWS Outposts supports first-generation (c5, m5, r5, g4dn) and second-generation (c7i, m7i, r7i, c8i, m8i, r8i) instance families, all EBS-optimized. These instances provide dedicated bandwidth to EBS I/O, reducing traffic contention. The maximum sustained performance of an instance is given by its IOPS and throughput specifications, but the lowest-performing subsystem (instance or volumes) will limit the overall. For example, an r7i.12xlarge can reach 60,000 IOPS and 1,875 MB/s throughput, but if EBS volumes do not meet those values, the instance will be underutilized.
Regarding EBS storage, first-generation Outposts racks support only gp2, while second-generation adds gp3. gp2 provides up to 16,000 IOPS per volume (3 IOPS per GiB) and a maximum throughput of 250 MiB/s. To achieve 16,000 IOPS with gp2, a volume of at least 5,334 GiB is needed. In contrast, gp3 offers the same maximum IOPS but with throughput up to 1,000 MiB/s, requiring only 32 GiB to reach 16,000 IOPS, as IOPS is provisioned at 500 IOPS per GiB. This provides a significant cost and efficiency advantage for database or data-intensive applications. At Q2BSTUDIO, we design architectures leveraging gp3 on second-generation racks to reduce provisioned storage without sacrificing performance, integrating AI solutions and AI agents to automate performance analysis.
To calculate required IOPS, first select an instance that supports your target, then size the EBS volumes in size and number. When using multiple volumes, total IOPS must equal or exceed the instance's maximum. With gp2, multiple large volumes are needed; with gp3, small volumes suffice. For example, for 60,000 IOPS on an r7i.12xlarge: three gp3 volumes of 32 GiB (16,000 each) and one of 24 GiB (12,000), totaling only 120 GiB, versus over 20,000 GiB with gp2. Throughput is calculated considering gp3 offers 125 MiB/s baseline and can be increased up to 1,000 MiB/s at a rate of 0.25 MiB/s per provisioned IOPS. For maximum, an 8 GiB volume configured with 4,000 IOPS suffices.
In database environments like RDS (SQL Server, MySQL, Oracle, PostgreSQL) on Outposts, IOPS are limited both by the database engine and storage. SQL Server, for example, supports up to 16,000 IOPS. To achieve this with gp3, a 32 GiB volume is required along with an instance like r5.4xlarge (18,750 baseline IOPS). On second-generation racks, r7i.4xlarge instances offer 20,000 IOPS, but the engine becomes the bottleneck. Q2BSTUDIO recommends continuous monitoring via CloudWatch, using metrics like VolumeReadOps, VolumeWriteOps, and latency times to adjust provisioned capacity. It is also advisable to use AWS Fault Injection Service to run controlled performance degradation experiments and validate resilience.
Cybersecurity also plays a key role: on Outposts, EBS encryption is enabled by default with AES-256 and AWS KMS keys, with no performance impact. Additionally, at Q2BSTUDIO we implement IAM policies to restrict who can create, modify, or delete volumes, as well as CloudWatch alarms to prevent unauthorized users from suppressing critical alerts. We combine these practices with cybersecurity and pentesting services to protect hybrid workloads.
In summary, architecture for IOPS and throughput on AWS Outposts racks requires a careful balance between instance, storage, and monitoring. At Q2BSTUDIO, as a custom software development company, we offer solutions integrating cloud AWS/Azure, artificial intelligence, AI agents, cybersecurity, and BI/Power BI to ensure your critical applications run at peak performance. Contact our team to design an architecture tailored to your specific needs.
