I recently returned from this year’s VMware Explore conference in Las Vegas. My trip was longer than just the conference, as I was fortunate enough to be selected to participate in the CTAB (Customer Technical Advisory Board) meetings that took place on Saturday and Sunday ahead of the conference. While the information around this, the presentations and meetings for these two days are subject to an NDA, what I can tell you is that if you ever get an invitation to attend this portion of the conference, you should accept it. If you are not familiar with CTAB, it is where customers are invited to meet with program managers, developers, engineers, and executives to discuss the good, the bad, and the ugly around the VMware ecosystem, and what can be done to improve it. This also gives us the opportunity to make recommendations about additional functions and features for the product. This information is taken seriously and is used to improve the quality and functionality of the product over time. It is time well spent, and there is a lot of work that goes into the product and great things are coming.
Overall, the conference did not take up as much space as in the past and was more consolidated. This was due to ongoing construction at the Venetian that limited the space available. A lack of space impacted the reason fewer submitted sessions were accepted this year. However, there were still a lot of very informative sessions to attend. Unlike previous conferences, the first day did not start with a keynote. That was later in the day and renamed the “Plenary Session”. So, it was straight to my scheduled sessions after breakfast.
Hardware
In today’s market, hardware and cost were the topics of several sessions. One of the key things to note is that Broadcom has made it where most of the hardware supported on vSphere 8 (or VCF 5.x) is also supported with VCF 9.x. With today’s cost of memory, CPU, and storage this was a welcomed announcement. This means that organizations can either move to or upgrade to VCF 9.x without the need to purchase additional hardware. This alone will save organizations millions of dollars. There may be some rare situations where some hardware will not qualify, but this should not be a common occurrence. One of the sessions where this was really demonstrated was in a session presented by John Nicholson on Wednesday titled “10 Tips for Reusing Existing Hardware for VMware Cloud Foundation 9 Deployments”. In this session, he covered changes that were made to the product to make this possible. As well as steps customers can do to minimize costs. These could include adding NVMe drives to existing hardware and enabling memory-tiering to supplement the need to add memory to systems. He also suggested maybe looking past the traditional OEM’s and that non-traditional ones seem to have 78% lower pricing on some hardware components.
Security
Numerous sessions covered enhancements in security that have been made by Broadcom when it comes to VCF 9.x, especially in this age of AI-driven threats and attacks. In one session, the evolution of AI threats was equated to the development of siege cannons against castle walls. What used to seem nearly impregnable has now been made much easier. Security has now been prioritized over anything else. Work has been done to harden the entire stack, from the VMware host kernel all the way up. Everyone hates patching, it can require downtime, break things, validations can be slow, and rollbacks difficult. Now, there is quick and live patch (although not possible in all situations with ESX) where no host maintenance mode or reboot may be necessary. vCenter can now also be updated and patched in less than 5-10 minutes rather than an hour or more.
Further up the stack, vDefend can not only enhance protection of east-west traffic with microsegmentation, IPS\IDS, and other features, but there is now virtual patching available. Have a CVE, but no patch available from a vendor, vDefend can be configured to protect critical workloads from being exploited until a patch becomes available. In addition, threat actors have become harder to detect due to using valid credentials and tools. However, now with vDefend and Ops, all traffic is logged for analysis. Illegitimate users will now show up in the data flows and when combined with automatic actions from IPS\IDS, are blocked before they can cause any harm. These are features definitely worth checking out.
Automation
VCF Automation was a big topic of the conference this year. What does automation provide you and why is it important? First, it provides a self-service portal for other users in your organization, such as developers. The automation allows them to go to the portal, configure a system for their needs, and then deploy it. This saves the administrator’s time by not having to go through the process themselves.
Besides a self-service portal, the automation is built-in to VCF 9 not only supports virtual machine workloads but also includes the integration and management for both AI and K8’s work. One session I attended was more of a tutorial and demo session walking through setting this up in an organization called End-to-End Automation Provider, Organization, Setup and configuration. This session walks through the steps necessary to set up VCF Automation. While this was done in a demo, it showed how this could be set up in about an hour, including multi-tenancy, regions, organizations, etc. This includes external sharing and access to external content libraries. This showed how VCF 9 has a true cloud experience in an on-premises solution.
VCF for Artificial Intelligence Workloads
There were many references to tokens during the week. These includes tokenomics, tokenization, etc. Basically, referring to the cost of running AI models in cloud-based services. Over the past year, there has been a lot of work put into VMware Cloud Foundation and its ability to efficiently run AI workloads utilizing Private AI Services. This service is used to deploy and govern both AI models and agents via containers. The benefits of using VCF for AI versus a cloud provider include some of the following benefits:
- Data Privacy and IP Protection
- Flexibility
- Not only sovereignty, but control over organizational data
- Better cost optimization
- Higher performance inference by being close to the data and utilizing pooled GPU’s
- Both audit and legal compliance
All the above helps, and organizations control cost. There have been numerous news stories earlier in the year about organization burning through their entire budget\tokens with months left in the year.
Using Private AI Services via API’s to create Infrastructure-as-Code can also save organizations over manual workflows and improve consistency. This is because traditional environments can have configuration drifts between systems where those deployed via API services are identical. In addition, deploying large model weights could create system constraints and issues causing network bottlenecks. Infrastructure-as-Code deployments can automate full-stack, flexible runtime environments that are more efficient and less likely to cause resource constraints.
In general, most workloads do not need the latest in frontier AI models. Using VCF for AI and running these workloads on-premises saves what organizations spend on tokens.
New Product Announcements
As in previous years, there were announcements of both new and coming products and services regarding VMware Cloud Foundation during the afternoon plenary (aka keynote) session. Here are the new features and services announced:
- VMware AI Factory
- This will be one platform to go from bare metal to model quickly, making it your model on your infrastructure. Currently, this solution is powered by AMD and Intel, along with both Dell Technologies and Cisco. However, additional vendors are constantly being added. This platform also supports the latest NVIDIA B300 Blackwell GPU’s, and with the announced partnership with Metalsoft, deployments to bare-metal servers are easily managed.
- TrueSource
- This will be the de facto protection for AI workloads using Spring Enterprise and its dependencies. This will create trusted artifacts and data services, even open sources.
- Agent Sandbox
- With VM’s and Kubernetes running AI workloads, this will provide guardrails with the use of explicit routes, protections from credential leaks, and agent build packs for prebuilt and third-party agents.
- Governed Marketplace
- This marketplace will provide published agents for sub-agent orchestration, improved agent performance, and controlling both short-term and session memory. This will also control access to various services.
- AgentMinder
- This will provide identity agents with rules and controls for each one. The product is currently planned for Avi Load Balancer and vDefend. It provides observability into how agents talk to each other and the entire ecosystem.
Overall, the conference provided a week of knowledge, announcements, peer-to-peer conversations. The construction may have reduced the size of the space available and reduced the number of accepted sessions, but I believe that overall, the conference was a success and enjoyed by all. It was announced that next year, the conference will take place not only earlier in the year, but at a new location, Resort World in Las Vegas. So, start planning now, get your proposals ready, and prepare for VMware Explore 2027, now happening May 3-7 in Las Vegas.
Leave a Reply