608B71FC-006A-4934-A643-7D9BA9340450Blog

Transform Hub, welcome to the Scramjet Family!

blog__author-img
Michał Czapracki
CEO at Scramjet, Data Streaming Expert.
23F2E8CD-3026-46A5-86CC-D13114F7176E425AE875-B1A1-4EA1-8529-075D08DA0BB1

24 January 2022

Scramjet Transform Hub (STH) is designed to drastically improve and simplify what has been possible with the Scramjet Framework, giving developers even more tools and freedom in the Cloud. Unlike the current Serverless systems, Scramjet Transform Hub gives you an opportunity to process data from different sources and in multiple scenarios. You can not only separately process small events, but also lists of local files, large files, pull streams from streaming APIs and so on. The choice is yours and STH adapts to your use case natively.

We looked at the competition and compared their solutions to our newly launched STH. This is a lot to process, but most importantly STH offers best-in-class performance metrics at an unrivalled cost while being simple to deploy, in just 4 steps! If Scramjet Transform Hub sounds interesting to you, go and check it out – we believe you’ll be able to unlock great new use cases with STH.

See how to get started on the Scramjet GitHub Repository.

Feature comparison across major STH competitors

Performance metrics comparison across major STH competitors

MerticLambda, Azure Fn, GCFOpenFaaS, NuclioKafka, Kinesis, Azure EHScramjet Transform Hub
Data size limit6 MBRAM1 MBUnlimited
Inter-step processing latency (95th)100 ms100 msN/A~ 10 μs
Maximum processing time15 min1 hN/AUnlimited

End-to-end latency level (w/o net). How much library (frameworks, caches, buffers) memory is needed to sustain operations of an 8 step workflow processing 8 concurrent items (compared to 1x).

64 x64 x8 x1 x

8 x 8 step external orchestrators. How many additional running systems are needed to sustain operations of an 8 step workflow processing 8 concurrent items.

9 x9 x9 x1 x

8 x 8 step serialization operations. How many times the data needs to be serialized and deserialized in an 8 by 8 scenario.

128 x128 x128 x16 x

Pricing model comparison across major STH competitors

MetricLambda, Azure Fn, GCFOpenFaaS, NuclioKafka, Kinesis, Azure EHScramjet Transform Hub
Unit of scaleEventDataMixedN/A
Timescalesecondssecondshrs / daysN/A
Pricing model complexityV-ComplexOpen SourceComplexOpen Source

Scenario low throughput monthly: 0.5 requests per second + 500 ms processing + 512 MB RAM + 10 kb output on average

$1,11Servers cost = $ 300N/AFree + servers cost = $100

Scenario generating data for ML 10 GB of items: 500 ms processing + 512 MB RAM + 100 kb output on average

$1,34N/A

Scenario large data output: 0.5 requests per second + 1000 ms processing + 512 MB RAM + 1 MB output on average

$113,82N/A

Scenario full month of ops: 1 requests per second + 1000 ms processing + 512 MB RAM

  • 100 kb output on average
$26,57N/A

Scenario 50% baseline + 8x scale: Like full month but with 8x peak for 4 hours a day

$60,08N/A

Scenario 1 TB of i/o ASAP: Best case for processing 1TB of input and 1TB of output data (with additional services when needed)

$154,13$159,65
Project co-financed by the European Union from the European Regional Development Fund under the Knowledge Education Development Program. The project is carried out as a part of the competition of the National for Research and Development: Szybka Ścieżka.