FlexPod is a converged infrastructure option developed by NetApp and Cisco. FlexPod is available in streamlined and simplified Express version, and the FlexPod Datacenter edition for enterprise private clouds, VDI, scale-out infrastructure, or software defined data centers.
N/A
NetApp ONTAP AI
Score 10.0 out of 10
N/A
NVIDIA and NetApp collaborate to deliver ONTAP AI, a converged infrastructure stack that incorporates
NVIDIA DGX A100, a 5-petaflop AI system, and NVIDIA Mellanox high-performance Ethernet switches to unify AI workloads, simplify deployment, and accelerate return on investment.
FlexPod is great for mid to large size companies, where the flexibility and depth of a traditional SAN and high performance servers is required. For smaller companies, it might make more sense to go with a hyper converged solution such as Hyperflex or Nutanix (both of which can still run on Cisco UCS servers, but would not be making use of Netapp storage), to meet the requirements in a smaller footprint.
If you want to deploy your AI system with precofigured solution and validated architecture and also want to scale your system, NetApp ONTAP AI is go to product. If you want to build not so large pipleline for analytics, or an ai system that dont require complex computation then its not recommended.
KVM control of the blades still requires Java. Avocent is using HTML5 now, and it would be nice if the KVM console for these UCS blades could too.
Price - Like any Cisco product, there are cheaper options. They aren't nearly as fully featured, but at times, it would be nice if UCS could be a bit cheaper.
Being a Cisco shop already it was a natural fit to go with a solution based on their technology. Then add in VMware which we are were already using - that was two out of the three for the FlexPod. Then factor in NetApp's de-duplication technology and flexibility in configuration and options and it's a match made in heaven. Not to say we haven't had some stumbles and some issues but it works and it works well.
1) Microsoft Cognitive Toolkit (Formerly CNTK) dont have vast functionalities they are limited to set of rules. 2) They dont support all in one. I.e NetApp ONTAP AI I can analyze, inference and compute on simgle node or system. 3) pricing of Microsoft Cognitive Toolkit (Formerly CNTK) is higher than NetApp ONTAP AI
FlexPod has allowed our team to be extremely quick to resopnd to new VM build requests. The amount of RAM, CPU and backplane offered by B series blades allows us to go with very high VM density. The quick deployment of service profile templates also means that when we have to add new hosts, it is done quicker.
FlexPod and the service profile portability (along with VMware) has allowed for full upgrades and migrations from M1/M2 series hardware to M4 hardware without any downtime or outages to the clients. Blades can be moved into new chassis, or service profiles moved to new generation blades, with no impact to the customers' ability to work.
