What is our primary use case?
Our primary use case for vSphere is not a primary use case, because we actually offer a pretty wide breadth of services. Our key use cases revolve around hosted private cloud, as well as being the underpinning virtualization platform for our multi-tenant vCloud Director based cloud.
We don't use VMware cloud on AWS.
How has it helped my organization?
vSphere has improved our organization by allowing us to deliver rock solid stability to our clients in a cost competitive fashion. The industry has moved far beyond bare metal infrastructure, other than for very specific us cases. As an operator of mission-critical applications on behalf of our clients, we chose vSphere because we needed the operability we get from features like vMotion, the stability that it gives us, and the ability to run pretty much any workload.
We host infrastructure for a very large number of clients. In many cases, we're running all their mission-critical applications in our data centers on top of vSphere. So, there is no single industry vertical. However, for each of our clients, we are their operator, and this is their mission-critical infrastructure.
When I think about the performance aspects of vSphere, we've been using it since before there was vSphere. We were actually a very early partner of VMware. I've been with NaviSite for a very long time, and I recall doing a VMware GSX Server deployment, from a number of years ago.
When I look at the performance aspects, I've definitely seen a reduction over versions from the virtualization penalty. This has been significantly reduced over the years. The size limitations of VMs, number of CPUs, amount of memory which can be allocated, and amount of storage which can be allocated are no longer of practical consequence. So, the monster VM that we talked about over VM Worlds of three to five years ago, they're here to stay, and those limits are no longer practical impediments to virtualization.
What is most valuable?
- The most valuable feature of vSphere is vMotion, because it rocks. It radically changes the way we think about how we can operate a large infrastructure, and notably, in terms of proactive maintenance.
- The second biggest feature is HA, because complexity around IT resilience is a difficult problem to solve, especially at the application level. Therefore, being able to rely on the infrastructure to provide a 90:10 or 99:1 rule is more than enough resilience for most applications, and getting that directly from the infrastructure is fantastic.
These features are useful day-to-day, because we operate a very large number of single-tenant private ESX deployments, managed by vCenter, as well as VCD-based public cloud. Frankly, with hundreds and thousands of hosts under management, there's no way we could operate that infrastructure without the use of vMotion. The ability to migrate those workloads to free up the physical infrastructure for maintenance activities, patching, BIOS updates, etc., is a critical requirement to operate.
An important vSphere feature from a security perspective is VM encryption. As is the right thing to do in this day and age, security needs to be the number one concern for any IT operator. While there are security solutions which can be delivered at the physical, hardware layer, they don't necessarily address all of the requirements from an encryption perspective. Being able to have VM-centric, VM-level encryption is a great feature of vSphere.
What needs improvement?
As with any piece of technology (hardware or software), there's always room for improvement. vSphere is incredibly mature from a core feature and function perspective. As we continue to push mission-critical workloads into vSphere, and those workloads are not readily protected at the application layer for availability, continuing to increase the size limitations on FT-protected VMs would be a great advance.
vSphere management has evolved over time. It's inherently complex. Operating a large virtual infrastructure is not an easy task for anyone. That's why certifications, such as VCP exist, because you have to have the right skill set to operate the environment. As the product evolves and starts to take advantage of things, like DRS, workload placement becomes less of an issue for humans to worry about, because the system takes care of it for you. Of equal interest is SDRS, storage management and storage placement, as historically, it was one of the most challenging things to mange in a large production VMware environment. With SDRS, we've actually seen our need to babysit it and manage it as a human go way down.
What do I think about the stability of the solution?
vSphere has been very stable. It would be where it is in the market overall if there were any sense of instability. No software nor hardware is perfect, so really it comes down to the failure rate that we see running workloads on vSphere. Is it significantly, materially, measurably different than running those workloads on bare metal? I would say absolutely not.
Equally important is the stability better because, when things happen, hardware is lost. In response, VMware HA automatically restarts those workloads and the effective downtime is radically minimized. This is compared to what it would be for a human response.
What do I think about the scalability of the solution?
Scalability on vSphere has always been important for us, because of the scale at which we operate. We had a client, who maxed out under the VMware 5 limit of 32 hosts per cluster. So, it has been great to see the continued improvements in scalability. At the VM level, the limits are no longer practical impediments. Now, at the VMware cluster level, we're also seeing sizes which can operate pretty much any large client environment.
How is customer service and technical support?
We've had to use vSphere and VMware tech support on a fairly regular basis, but not because there are fundamental flaws in the platform. Things happen. Client environments are complex, and in some cases, the interoperability with other third party products requires engagement with support. We have found the engagement able to solve our problems pretty much all the time.
How was the initial setup?
I'm not directly involved in the day-to-day operations of our vSphere environments, but we stand up private vSphere-based clouds on a fairly regular basis. We manage those on a go forward basis in terms of patching, upgrading, etc. Deploying vSpheres is pretty easy. The biggest feature that has made that easier, as compared to three or four years ago, is the vCenter Server Appliance. Its ability to deploy the management plane as a virtual client and bootstrap an ESX environment. That's a big step forward.
What was our ROI?
- Compared to deploying traditional infrastructure models, like bare metal, and the ability to virtualize and maximize the utilization of the physical infrastructure speaks well for ROI.
- In today's market, agility is the new currency. Without virtualization, and vSphere in particular, we wouldn't have the level of agility in the business that we have today. Frankly, it's needed by pretty much any industry. Regardless of whether you're technology-centric or not, you are a technology company.
What other advice do I have?
If I had to give a rating of one to ten for vSphere, I would give it a nine. No software nor hardware is perfect, but vSphere is good. That's why I would say a nine. There is still some room for improvement, like larger FTVMs, continued evolution, and keeping pace with the scalability of underlying physical infrastructure.
For somebody looking to evaluate a virtualization platform such as vSphere or any of its competing open source solutions, like KVM or other virtualization platforms, one of the key considerations is to look at TCO. vSphere may seem expensive upfront, and there may be some sticker shock there, but if you look at it over the long-term and from a human capital perspective to operate the platform over a period of three or more years, the manageability of vSphere drives the total cost of ownership way down.