Selecting a Frame Deployment Model for a Global Design Firm
A design firm has 220 employees in London, New York, and Singapore. 80 use GPU-accelerated CAD applications; 140 use Office and browsers. They want to move from on-prem workstations to Frame. They have an existing Nutanix on-prem cluster in London with spare capacity and a 10 Gbps internet link. Data residency requires EU employee data stay in the EU; US and APAC data can be in respective regions. Users need persistent personal drives and occasional persistent desktops for contractors. The firm wants to minimize capex, but the CFO is open to reserved capacity. Current internet latency from Singapore to London is 180 ms. They ask whether to use Frame public cloud, Frame on-premises, or a hybrid. They also need SSO with Azure AD and MFA.
1. Recommend a Frame deployment model and justify it using latency, data residency, and cost constraints.
Read the worked answer
A hybrid model is the best fit. Run Frame on-premises in London on the existing Nutanix cluster for EU CAD users and EU office users, keeping data in the EU and taking advantage of spare capacity. For New York and Singapore, use Frame public cloud in the respective region rather than backhauling to London. The 180 ms Singapore-to-London latency is unsuitable for interactive CAD; in-region public cloud should keep round-trip latency in the range where Frame streaming remains usable. Data residency is satisfied by keeping EU profiles and storage in London, US data in a US region, and APAC data in an APAC region. From a cost perspective, use reserved capacity for the predictable baseline of GPU CAD users and on-demand or auto-scaled capacity for peaks, contractors, and office users. This avoids building a second on-prem site while preserving EU sovereignty and acceptable user experience.
2. Design the identity, profile, and persistent-storage approach, including the contractor requirement.
Read the worked answer
Use Azure AD as the identity provider with SAML SSO into Frame, and enforce MFA through Conditional Access. Map groups to Frame pools so that EU CAD users, office users, and contractors receive different entitlements. For profiles, use FSLogix profile containers where non-persistent desktops are used, and place EU containers on Nutanix Files in London; use regional cloud file storage for US and APAC pools. For contractors who need persistent desktops, create an isolated persistent pool with time-bound accounts, MFA, restricted clipboard and drive redirection, and no access to EU data unless explicitly approved. Persistent personal drives should be region-specific and encrypted. Do not let contractor desktops share storage with regulated EU CAD data. Document retention and backup for each region, and test profile size and logon performance before moving all users.
3. What GPU capacity strategy would you use for the 80 CAD users, and how would you validate it?
Read the worked answer
Model concurrency rather than assuming all 80 CAD users work simultaneously. If peak concurrency is around 50-60, reserve GPU capacity for the baseline and use auto-scaling for the remainder. On-premises in London, use supported NVIDIA GPUs with Frame on Nutanix AHV and dedicate GPU hosts to CAD pools. In public cloud regions, select GPU instance types that Frame supports for the CAD application, and keep the same vGPU profile class across regions to simplify image management. Do not overcommit GPU memory; CAD and 3D workloads are sensitive to framebuffer pressure. Validate with a pilot that measures session latency, frame rate, and application launch time from each region. Use Frame monitoring to track session count and GPU utilization, then adjust reserved versus on-demand capacity monthly.