Kaminario playing 3D flash chippery doo-dah with its arrays
Now it's just looking for a little TLC
A connection is made
El Reg: Do you see a need for public cloud connectivity for K2 arrays, to provide data protection or other facilities?
Fienblit: We believe that all-flash arrays deliver the best mixture of performance and cost in a private cloud and cloud hosting environments.
We also see the benefit of utilising the cloud in some services which are less related to the active data and we are working in collaboration with other partner vendors for delivering solutions of connecting K2 arrays to the public cloud.
We are also examining native K2 functionality to connect directly to the cloud, allowing our customers to take advantage of our simplicity and easy management, also for public cloud usage. That said, for replication purposes it will mean very long RTOs, which is not something many of our customers are looking for.
El Reg: Do you see a role for Kaminario products in Big Data-style applications, such as analytics?
Fienblit: Analytic is one of the biggest trends in IT and customers are looking to scale analytics capabilities in a very cost-efficient way. This is the most common use case of K2 systems across all industries. We see that 40 per cent of the use cases that are running on K2 systems are analytics loads and in practice, we see that customers are running real mixed workloads of analytics and online processing on the same data set.
Kaminario architecture has a significant differentiation in this area, as we implemented an adaptive block size algorithm that enables us to achieve 2-4 times higher bandwidth-per-node compared with our competitors that implemented a content address mapping algorithm.
This immediately translates to cost, as we can meet the performance needs for analytic environments using less hardware compared with our competitors. In practice, we see that the average IO size in our systems is 43,000 and that 70 per cent of the read bandwidth across all our customers is 64,000 and above.
To summarise the roadmap hints here, we can expect Kaminario to introduce 3D TLC NAND SSDs in a few months, possibly by early 2016. It may well have MKC and TLC tiers in its K2 arrays, and add a cloud back end for low-access-rate data, either directly or through partners.
The thrust of its strategy is to take advantage of flash improvements to provide more cost-efficient all-flash arrays than the competition and so extend their use case.
We wouldn't be surprised if it was looking at another SPC-1 benchmark run to try and reclaim the flash cost-efficiency crown from HDS. ®