(Burstable Performance Instances)
AWS has been coming up with various EC2 instance offerings that are constantly evolving and staying on top of the cloud market. These EC2 instances form the backbone for the entire AWS cloud ecosystem. Therefore, it becomes extremely important to explore the details and uniqueness of each instance family for better decision making.
In today’s blog, there is an emphasis on General-purpose instance categories, especially the Burstable instances which help in both cost and performance optimization, if chosen wisely.
Most of the General-purpose workloads fall under 3 instance series viz. A1, M4, M5, M6, T2, T3, T4.
While every workload is different in its own way, in terms of the workload size & CPU utilization patterns, T instance family is unique wrt a wide range of instance size starting from nano to 2xlarge (smaller end of the scale) compared to M instance family.
Moreover, T instances have gained more popularity in recent years because of the “Burstable performance” nature, due to which customers have not only got benefited from considerable cost savings but also had experienced the same performance with that of M4 or M5 instances as long as CPU% is within the range of “Breakeven threshold” (will discuss later in the blog series)
As per the analysis done by AWS on the applications and instances that are running on General Purpose workloads, it is been observed that a broad set of General Purpose instances use low CPU utilization and only a fraction of the instances are using high CPU utilization. This was the driver for AWS to come up with Burstable instances.
Most popular Burstable ‘T’ Instances — T2,T3, T3a,T4g
Comparison of T2,T3,T3a and T4g Instance types
By design, the instance is built to have a moderate CPU usage that is included in the offering. If you want to sustain high CPU usage, you need to pay as you go on-demand price for that incremental CPU usage. It is very much ‘pay for what you use’.
Quick facts for consideration
- T3 offers 30% improved price to performance compared to previous generation T2 instances.
- T3 is based on Intel Xeon Platinum 8000 series (Skylake) processors with a core Turbo clock speed of 3.1 GHz/sec.
- T3a is based on AMD EPYC 7000 series processor with a core Turbo clock speed of 2.5 GHz/sec.
- The default mode of CPU burst setting is “Unlimited” for T3 & T3a Instances. It can be changed to Standard and vice versa at any time.
- T3 generation instances come with twice the # of vCPUs on smaller instance sizes (nano, small) when compared to T2.
- Intel Hyperthreading is enabled by default so that customers get the incremental performance that is enabled by the Intel processor chips.
- T3 instances are built on EC2’s AWS Nitro system, which gives more access to the CPU power and higher network and storage performance.
- There is an improved network performance of up to 5 Gbps at burst.
- There is an improved EBS performance offering up to 1.5 Gbps on smaller sizes and 2.05 Gbps on larger sizes.
- T3 instance types support only EBS storage and not Local storage.
- T3 is 13% cheaper than M5
- T3 is 17% cheaper than M4
- T3 is 10% cheaper than T2
- T3 is 30% better price to performance than T2
- T3a is 10% cheaper than T3
- T4g is up to 40% better price to performance over T2 and up to 20% over T3
- CPU Credit Price is 5 cents per vCPU-Hour for Linux, RHEL & SUSE & 9.6 cents per vCPU-Hour for Windows
Lot of customers whose general-purpose workload’s average CPU utilization is idle to low to moderate & at times occasional periods of high CPU activity have got benefited by migrating their existing workloads on M instance types and some old generation C type instances to Burstable T type instances and getting benefited with the performance they need at a lower cost.