2022 cloud report comparing cloud performance of Amazon, Google, Microsoft will be released



Cockroach Labs, a software company that develops commercial database management systems, has released the 2022 Cloud Report, which compares cloud performance with benchmarks.

2022 Cloud Report | Cockroach Labs

https://www.cockroachlabs.com/guides/2022-cloud-report/

Cockroach Labs has been conducting cloud benchmark tests every year since 2017, and in 2022, users asked a lot of questions, 'Which of the small but large number of nodes or the large but small number of nodes will be used? The survey focused on 'Is it good?' Therefore, benchmark tests were conducted on three cloud services, Amazon Web Services (AWS), Google Cloud Platform (GCP), and Microsoft Azure, mainly on four axes: OLTP performance, CPU performance, storage I / O performance, and network performance. I did.

First, let's compare the OLTP performance of the instance types used by each cloud service. For large instance types, GCP t2d-standard-32 with AMD Milan processor is at the top. N2-standard-32 with Intel Ice Lake is in 2nd place, and AWS m6i.8xlarge is in 3rd place.



Even for small instance types, t2d-standard-32 is still at the top. The result was m6i.2xlarge in 2nd place and n2-standard-8 in 3rd place.



Cockroach Labs says, 'Based on OLTP performance, you can get a competitive solution in terms of price / performance regardless of which of the three cloud services you choose.' In the graph below, red is GCP, blue is Microsoft Azure, orange is AWS, dark colors of each color are reserved, and light colors are on-demand. The price is the price per

TPM with an annual contract.



Next, the result of CPU performance is as follows. AMD Milan dominated the top five, with GCP t2d-standard-8 in first place, GCP t2d-standard-32 in second place, and Microsoft Azure standard_E8as_v5 in third place.



The storage performance results are below. The left side is a small instance, the right side is a large instance, and the median IOPS is shown separately for writes (dark) and reads (light). 'We were able to get at least spec performance with all of these storage options,' said Cockroach Labs.



Also, even with a relatively small amount of persistent block storage, the cost of running a particular workload is far more impacted by the cost of storage than the cost of an instance. Unless your workload requires a lot of IOPS or very low latency, Cockroach Labs says it's best to choose a mid-tier storage option (GCP pd-ssd or Microsoft Azure premium-disk).

Network performance measures such as the average cost of running a single throughput in a cross-regional network of each instance type. The graph below shows the performance and cost of running for 12 minutes each on AWS, GCP, and Microsoft Azure. Cockroach Labs said, 'GCP was overall superior in terms of both performance and cost.' In addition, he said that in the future, we will deepen our analysis of such 'hidden costs' because these costs can affect total costs more than the costs of the instance itself.



Finally, when asked which is better, small or large instances, Cockroach Labs replied, 'In terms of TPM per vCPU, small instances are significantly better.' Tested with OLTP and CPU benchmarks, eight of the top 10 instances are small instances, with overall performance and average performance superior to large instances.



You can download the report with other detailed data by clicking 'Download the 2022 Cloud Report' from the Cockroach Labs link below and filling out the form.

2022 Cloud Report | Cockroach Labs

https://www.cockroachlabs.com/guides/2022-cloud-report/



in Web Service, Posted by log1p_kr