Design a VMware environment with CPUs, storage in mind
A solid infrastructure is critical to smooth operations in the data center. Admins designing a VMware environment should think about CPUs, storage and reference architectures.
If one part of a complex data center doesn't work properly, it can have a devastating effect on operations. However, you can design your VMware environment to prevent such snags. Consult reference architectures, and think carefully about which CPUs and storage systems you select.
You might take design for granted if you focus heavily on support and installation, but building a solid foundation will help ensure production workloads run smoothly. Poorly planned data center architecture, on the other hand, can leave you scrambling as your environment grows and lead to all kinds of performance issues.
When you design VMware architecture -- or any IT architecture -- you won't find a one-size-fits-all method.
There isn't a one-size-fits all approach to IT architecture. Your company and customers are unique, and your budget, security and performance requirements are, too.
You should tailor your VMware architecture to meet the specific needs of your organization. However, don't dismiss other reference architectures; they often contain helpful information. Your data center can incorporate a collection of design aspects from several sources, but keep your organization's limitations in mind.
Decide on CPU core count and speed
When you're choosing a CPU, you should focus on the needs of your workloads over your loyalty to a specific vendor.
If your workloads or VMs are mostly transaction-based, they might require more cores per CPU socket, which enables them to process more threads at once. If you run more concentrated workloads -- such as databases -- then you might want CPUs with higher speeds and fewer cores. If you run a roughly equal number of transactional and concentrated workloads, find CPUs that give you a healthy balance between core count and core speed.
Also, keep licensing in mind. VMware licenses per CPU socket, but Microsoft requires additional "core packs" beyond eight cores per CPU socket. This shouldn't be the only thing that drives your architectural choices, but you should factor in how costs add up with certain CPUs.
Maximize your host storage
You might overlook local storage if you use a storage area network or network-attached storage, but using local storage for hosts can have a profound effect on your overall environment. Some of the benefits of local storage include better performance, speed and control over security.
Local storage often comes in two forms: a spinning disk or solid-state drive (SSD) with RAID protection, and non-volatile memory express (NVMe) cards for high-speed access. Each type of storage plays a role in your data center.
Spinning disks and SSDs provide relatively fast storage with low latency, which equips them to handle both large and random workloads, but Serial-Attached SCSI-connected SSDs' connection protocol limits speed if you use them for caching.
You can use NVMe storage to set up vSAN or for host-side caching. However, if an NVMe card -- especially one used as your only form of VM storage -- fails, you might lose data.
Use shared storage wisely
VMware vSAN technology often costs less than traditional shared storage. Ideally, you should preserve shared storage for your VMs and other traditional workloads, but you can also use shared storage to house ISO images and templates.
You can remove such items from your primary shared storage to save money and space, although this means that not all hosts can access your templates and ISO images. You should keep backups of these files, and when you deploy them, deploy from a single host.
