It’s not that long ago that conventional wisdom dictated carving out a purpose-built place in your data center for hosting virtual desktops and apps. This sensible advice came from the realization that desktop and application workloads (and their users) can be largely unpredictable in how they consume CPU and storage resources. This has been especially true with virtual desktops incurring significant read and write IOPS, forcing storage admins to plan for worst case, and incur a heavy capital expenditure burden to protect for peak demand. Of course, that also meant lots of wasted CAPEX with idle spindles and flash being underutilized when the IOPS deluge subsided.
This kind of problem statement is becoming a relic of the hardware-defined data center, but does it really make sense where organizations are increasingly trying to enable cloud-like elasticity in how resources are delivered across multiple workloads in the datacenter? Now more than ever, IT architects want to have multiple, virtual workload domains that can be co-resident on the same infrastructure footprint, intelligently and dynamically balancing the allocation of capacity in response to the ebb and flow of application demand.
This kind of application centricity, instead of hardware centricity, is what’s needed.
To that end, SolidFire and VMware have partnered with Dell and Cisco to deliver two 
Agile Infrastructure Solution Designs. The SolidFire AI for Virtual Infrastructure is a pre-validated reference architecture, that intelligently consolidates multiple mission-critical workloads onto a single platform, based on SolidFire either Dell R630 compute with Force10 switching or Cisco UCS B200’s with Nexus switching. In the Dell RA, SolidFire SF2405 block storage is used, and in the Cisco RA, it’s SF3010.
There are some great proof points that go a long way to demonstrating the performance credentials that AI brings in support of workload mixing. The reference architecture validation process employed the following running concurrently:
- Boot storm of up to 500 VMware Horizon 6 desktops
- Two hosted Oracle OLTP database workloads running approximately 230,000 TPM
- MongoDB job running 70 concurrent threads
- Inline data reduction, including de-duplication, compression and thin provisioning, driving 13.5x efficiency rates across all three workloads
If you want to learn more about the Cisco UCS-based AI Reference Architecture for VMware, check out this solution brief. For more info on the Dell R630 Reference Architecture, check out this brief.
And if you want a solid discussion on best practices for building a winning solution for large-scale VDI deployments, check out this webinar presented by our own Tristan Todd, along with SolidFire’s Garrett Clark.
