VMware says server sprawl is back, and SmartNICs are the solution

More details emerge about 'Project Monterey' and its role driving infrastructure workloads into hybrid clouds

VMware has gone back to its roots to explain why the world needs to add a new hardware tier to datacenters.

Those roots lie in the scourge of "server sprawl" – a phenomenon that was rampant in the early 2000s when applications and hardware were tightly coupled, meaning that servers were specced to handle peak loads but hardly ever used their full power.

By abstracting server fleets into pools of logical resources and repackaging servers as virtual machines that could consume resources from those pools, VMware removed the need for tight coupling and allowed organizations to acquire and run fewer servers without compromising performance.

VMware nailed this tech just as the Global Financial Crisis swept across the world, and surfed it into the heart of enterprise IT. It's been there ever since.

Last week, VMware argued that servers have again started to sprawl and it again has the answer – this time in the form of "Project Monterey".

Virtzilla has been working on Project Monterey since at least 2019, when it showed off an Ubuntu VM running on a SmartNIC – a network card that includes a multi-purpose system-on-chip – and suggested such a VM would be a fine place to run network-centric workloads so that server CPUs were free to run applications.

The likes of Nvidia, Broadcom, Intel, HPE, and Dell have since bought into that idea (even if they're all following the hyperscale public clouds that have done it for years).

But software to make SmartNICs usable in a mainstream IT shop is not yet available.

VMware plans to change that with Project Monterey and claims one reason it will succeed is … server sprawl!

In promo material that has dribbled out in recent weeks, VMware argues that cloud-native applications are great, but also create a lot of east-west traffic that has increased the amount of scutwork CPUs need to handle. The company has also cited a paper that calculates the infrastructure requirements of some AI/ML workloads are so high that just a third of CPU capacity is left to run the actual AI/ML apps.

Organizations are therefore buying more servers to get the power they need. VMware reckons they’d be better off buying SmartNICs to make their current servers more efficient.

Enter Project Monterey. VMware is billing it as the way to deploy and manage server-saving workloads to SmartNICs with the vSphere and vCenter tools present in many IT shops.

VMware's new material also suggests Project Monterey has a role in hybrid clouds – a tantalizing prospect, given hyperscalers' SmartNICs are currently off limits to end users. The big clouds use them to keep their own houses in order.

If VMware has teamed with major clouds to offer SmartNIC as a service, that would be quite a change.

It's more likely that hybrid clouds in this context will come from the VMware Cloud Provider Program (VCPP), which has over 4,000 members that offer cloud services based on the VMware cloud stack. If VCPP members equip their servers with SmartNICs and deploy Project Monterey, they could offer nicely differentiated and efficient services that get a little closer to hyperscale efficiency.

The new material makes no mention of when Project Monterey will debut, but is clear it will be integrated into vSphere.

The documents also make frequent mention of the need for a new paradigm for IT infrastructure that integrates SmartNICs – and Project Monterey – as the harbinger of that change.

Ironically, VMware must share some responsibility for this decade's version of server sprawl. The success of its compute virtualization platform has seen servers replace special-purpose hardware like storage arrays, switches, and routers. The same servers that now need help from SmartNICs – and VMware – to make them work. ®

Similar topics

TIP US OFF

Send us news


Other stories you might like