![FlexPod XCS vs. Vblock [EOL] report](https://www.peerspot.com/assets/media/images/icons/pdf-80836ec5.svg){kind=icon}
![Vblock [EOL] Logo](https://images.peerspot.com/image/upload/c_scale,dpr_3.0,f_auto,q_100,w_64/nb70b5ansummme9cc5q5ymwyalbv.png)
![Vblock [EOL] Logo](https://images.peerspot.com/image/upload/c_scale,dpr_3.0,f_auto,q_100,w_30/nb70b5ansummme9cc5q5ymwyalbv.png){kind=small}
![Vblock [EOL] Logo](https://images.peerspot.com/image/upload/c_scale,dpr_3.0,f_auto,q_100,w_40/nb70b5ansummme9cc5q5ymwyalbv.png){kind=small}
The traditional UCS Blades do not take much storage internally. You would be challenged to create an HCI (Hype converged Infrastructure) solution on FlexPod / UCS or any other solution that pools internal storage. Now, with UCS X-Series, you can carve off an HCI solution, software defined pooled solution if you want. This was one area of improvement that I wanted to see and can now realize with the refresh of the Cisco UCS infrastructure. With modern modular infrastructure, RESTful API has been added, there are more integrations, ServiceNow and vCenter along with tighter plug-ins. There is cross-user interface launching, for example with Windows Admin Center. The solutions are using Ansible and Terraform for deploying infrastructure as code. All the improvements that I wanted from the last gen are here or coming. With modern workloads and GPU use on the rise, adding GPUs to modern modular infrastructure will have some pros and cons. Typically, you can add one or two GPU's to a blade with no or little trade off. With the UCS X-Series, if you are doing a GPU farm, then you may have to sacrifice compute blades in the front slots to put in a GPU tray / module. A chassis holds eight compute blades, but if you are adding a ton of GPUs, a single GPU tray or more will reduce your blade count by as many GPU trays you add. This is not just a Cisco UCS X-Series problem. It is an industry problem with modular infrastructure and one that I would like to see get solved! I am looking into one such solution, VMware BITFUSION where you can send CUDA requests over the network to a BITFUSION server with the results sent back to the requestor, early stages here and only scratched the surface thus far. With Cisco UCS X-Series, I would like to see the fabric interconnects built into the chassis instead of being external. With the fabric interconnects, the real footprint of UCS X-Series is 9U, where some of the competing solutions are 7U and have collapsed the network fabric into the chassis. This is another thing that I would like to see from Cisco, though, not really on the NetApp side of the fence, NetApp is solid storage.
