AWS SQS vs Memphis
This section describes the differences between AWS SQS and Memphis
Last updated
This section describes the differences between AWS SQS and Memphis
Last updated
Amazon Simple Queue Service (Amazon SQS) offers a secure, durable, and available hosted queue that lets you integrate and decouple distributed software systems and components. Amazon SQS offers common constructs such as dead-letter queues and cost allocation tags. It provides a generic web services API that you can access using any programming language that the AWS SDK supports.
Memphis is a next-generation messaging queue.
A simple, robust, and durable cloud-native message broker wrapped with an entire ecosystem that enables fast and reliable development of next-generation event-driven use cases.
Memphis.dev enables building next-generation applications that require large volumes of streamed and enriched data, modern protocols, zero ops, rapid development, extreme cost reduction, and a significantly lower amount of dev time for data-oriented developers and data engineers.
Benchmark
300K messages per second per station (queue).
60K-100K messages per second
Message Retention
Policy-based (e.g., 30 days)
Acknowledgment based
Data Type
Transactional, Operational
Transactional
Consumer Mode
Smart broker/Smart consumer
Smart broker/dumb consumer
Topology
Publish/subscribe based
Not supporting broadcasting. 1:1 producer to subscriber
Payload Size
Up to 15M
Max of 256KB
Batch size
No limitation
Max of 10 messages
Use Cases
Massive data/high throughput cases | Simple use cases
Simple use cases
Delivery Guarantee
At least once, Exactly once
At least once
Message ordering
Message ordering is provided via consumer groups. By message key, messages are sent to stations.
FIFO is optional
Message priorities
Unavailable
Unavailable
Message lifetime
Since station messages are kept on file/memory. This can be controlled by defining a retention policy.
Because SQS is a queue, messages are discarded after being read, and an acknowledgment is given.
Clustering
Active-Active
Unavailable
Performance
Scale-up, Scale-out
No control
Multi-region
Supported*
No
Multi-tenancy
Supported*
No
Read-replicas
Supported*
No
Data striping across nodes
Supported
Unavailable
*Available for Memphis cloud users
SQS uses a distinct, bounded data flow. Messages are created and sent by the producer and received by the consumer.
Memphis uses an unbounded data flow, with the key-value pairs continuously streaming to the assigned station.
AWS SQS is best for transactional data, such as order formation, placement, and user requests.
Memphis works great for transactional and operational data like process operations, auditing and logging statistics, and system activity.
AWS SQS pushes messages to consumers. These messages are removed from the queue once they are processed and acknowledged.
Memphis is a log. It uses continuous messages, which stay in the station (queue) until the retention period expires.
AWS SQS doesn't support multi-tenancy but through a lambda function, required to be code and managed by the user that acts as a router.
Memphis supports multi-tenancy using namespaces which offers a complete separation from connections, producers, consumers, security, dedicated dashboard, including node selection.
Some level of observability can be received by using 3rd party apps like Cloudwatch/Datadog/New Relic. To understand the full path of a message, it is required to use AWS X-Ray and add some headers to each client. Notifications can be achieved by building a dedicated event queue with lambda triggers. Some alarms and triggers must be defined over 3rd party apps to enable lag identifications and latency in real-time.
Memphis offers full Infra-to-cluster-to-data GUI-based observability, monitoring, real-time message tracing, and notifications embedded inside the management layer, including self-healing policies based on the defined events.
GUI
Yes
Yes
Schema Management
Yes
No
Wildcard consume
No
Yes
Stream Enrichment
Yes
Yes
Ready-to-use source/sinks connectors
Yes
No
Stream lineage
Yes
No
Data-Level Observability
Yes
Yes
Self-healing
Yes + Managed service
Managed service
Deduplication
Yes. Modified bloom filter
Deduplication interval of 5 minutes
Delayed queues
Yes. Atomic per message.
Yes. Not atomic, and per entire queue.
Dead-letter
Yes
Yes
REST Gateway
Yes
No
Consumer internal communication
Experimental
No
Production deployment environment
Kubernetes, Docker, Managed service
Managed service
Storage tiering
Disk, Memory, S3 for Archiving
Disk
Notifications
Slack, Email, More
With SNS and Cloudwatch
SDK support
Node js, Python, Go, .NET, Java, NestJS, and Typescript
C++, Go, Java, .NET, Python, node.js, Rust, Ruby, PHP
AWS SQS Client node: 1 x m4.2xlarge / 50 threads
Memphis Client node: 1 x m5n.8xlarge / 20 threads
100K messages = 100MB
0.16951 sec
5.88 sec
500K messages = 500MB
2.74 sec
29 sec
1M messages = 1GB
9.419 sec
58 sec
10M messages = 10GB
106.576 sec
588 sec
In this economic climate, costs are top of mind for everyone.
Total Cost of Ownership (TCO) should be a primary consideration when evaluating the Return on Investment (ROI) of adopting a new software platform. TCO is the blended cost of deploying, configuring, securing, productionizing, and operating the software over its expected lifetime, including all infrastructure, personnel, training, and subscription costs.
For this comparison, we define TCO as a combination of the following components:
Implementation: The cost of implementing a new streaming technology
Infrastructure: The cost of computing and storage, in this case on AWS
For the infrastructure cost comparison, we ran benchmarks to compare the performance of AWS SQS against Memphis.
Oftentimes, there is a misconception that cloud services are a turnkey solution. Here are some of the missing components that will need to be constructed when using AWS SQS -
Performance
Built-in. Automatic.
0 dev hours
Required to use threads.
130 dev hours
Multi-Cloud
Built-in. By design.
0 dev hours
Required to build abstraction to different cloud queues and APIs. 378 dev hours
Multi-tenancy
Built-in. By design.
0 dev hours
Required to build. Using different queues and/or tagging data.
63 dev hours
Monitoring and Notifications
Built-in. Ready-to-use slack notifications / Grafana / Datadog / prometheus.
12 dev hours + 2 DevOps hours
Required to build + use 3rd party open-source/paid tools.
126 DevOps hours + 49 dev hours
Runtime DLQ consumption
Built-in.
0 dev hours
Required to build.
30 dev hours
Delayed consumers
Built-in.
0 dev hours
Required to build.
20 dev hours
Cost
$1,000
Based on average dev hourly rate of $70.
$55,720 ($54,720 difference)
Requests per month
30,000 messages per second = 77,760 million messages per month
30,000 messages per second = 155,520 million requests per month
*Every Amazon SQS action counts as a request, including DLQ, publish, subscribe
EKS
Organizational.
None.
Nodes
3 x i3en.large (On-demand). $547
Included.
Client nodes
1 X M4.2xlarge.
$324.12
12 X M4.2xlarge
$3,889.44
Retention (Storage)
3 days X 10 kb message X 77,760M = 6Tb.
$1,737.52
3 days
Data Transfer between AZs
$2,866
$1,433.60
Licensing
Memphis Self-hosted enterprise licensing for partners. Flat.
$1,100 – $2,000
Included.
Cost
$6,574
Based on average dev hourly rate of $70 $56,655.60 ($50,081 difference)
Memphis.
Implementation
$100
$100
$100
Memphis.
Infrastructure
$4,330
$6,574
$9,909
Memphis.
Total
$4,430
$6,674
$10,009
AWS SQS.
Implementation
$4,643
$4,643
$4,643
AWS SQS.
Infrastructure
$35,709
$56,655
$99,516
AWS SQS.
Total
$40,352
$61,298
$104,159
Summary.
Implementation
x46
x46
x46
Summary. Infrastructure
x8.2
x8.6
x10
Summary. TCO
x9.1
x9.1
x10.4
Performance heavily relies on the client’s threads
What if you required to run on GCP for specific customer?
Not built for SaaS.
No multi-tenancy
Consumer-side delay queues
Monitoring and notification
Consumption from DLQ
