Cloud workloads with HPC, AI/ML in the cloud. Was launched at re:invent in 2017. WEKA is utilizing a data platform in order to support the major data driven solutions around HPC and AI/ML which helps fit in those areas

Now I have to admit I don’t know a ton about AI/ML pipelines when dealing with enterprise use cases so a picture is worth 1000 words

with the amount of steps all including data copy from the data learning from itself, you can truly see how data within these workloads is a pain to manage and keep.

WEKA is available in 4 cloud marketplaces including AWS, Azure, GCP, and Oracle cloud. WEKA is deployed via cloud formations or terraform and will live within the network cordoned off for it (VPC, Virtual networks etc.)

Personally I believe it’s a Hybrid world more than anything else. Most organizations I see are not just running things within one or two clouds, but also have on-premises locations and need to be able to connect them in their solutions. If only it was performant

oh it is. Worth mentioning the 2TB/s is in OCI and OCI is probably one of the most powerful cloud vendors out there.

It’s always great to understand the hidden cost to a solution, so I applaud WEKA for including this slide to help us understand what benefits are also hidden with adopting WEKA. WEKA has even created a guarantee for customers that they can cut the infrastructure bill in half for those customers (outcomes may vary).

LETS TALK NERD KNOBS!

What does WEKA look like in AWS?

Cluster of I3en servers with autoscaling, however you do not need to utilize bare metal instances in order to use WEKA. WEKA will assist customers in sizing and scaling within their environment and help them understand the total cost of ownership. This also includes networking as it’s a core piece of the network that needs to be validated with the storage needs.

The world is Hybrid, so the ability to replicate that data copy snapping to object storage is awesome.

When you need WEKA it stands up and runs the solution needed, but its able to Scale to zero and remove the infrastructure required for the data to be snapped, and then pushed back into on-premises in order to continue working on the data itself.

Very cool when thinking of bursting to the cloud.

If you would like to see this in action we were informed of this YouTube of the solution: https://www.youtube.com/watch?v=AZTzWKX-CrQ

My Thoughts:

I work with a number of incredibly smart individuals that work around AI/ML and the needs for GPU, Performant data structures, and other solutions. With WEKA I have been definitely shocked by the amount of depth they have gone into to fix the issues that their customers are find and helping them figure out how to address those problems with multiple avenues. First they work to make sure that what they are providing is performant and will be usable for customers that have the need of high IOPS data platform. Second, they work to provide a resilient platform that can be built with multiple DR and HA perspective so that it can be durable and maintain the customers solutions. Third, there are multiple form factors that will fit into the customers ecosystem and allow them to utilize WEKA in a way that fits their need, and their budget. Cost performance of WEKA is amazing with the ability to even scale to zero so that there is no consumption within the cloud. Finally, WEKA steps into the customers environment and works with them as a trusted advisor to help them consume their AI/ML solutions within the WEKA data platform, and help them evaluate and use the solution in the best way possible. With these standard ways of working with customers it ensures the customer is successful and that WEKA continues to grow from that into the future. I personally have seen the value to WEKA, and with the solutions they have presented see how they will keep their place in the ecosystem of AI/ML and HPC.

Interested in learning more about weka you can start at https://start.weka.io to see what you can do with the product.