I don't know how much I've updated the blog about this, but TECT is in the process of changing. Sort of.
When modern TECT started out, it was a single installation of Windows Small Business Server 2011 Standard directly on the metal of the TECT hardware. This had the advantage of working well and offering high performance, plus backups have been "sort of" easy, and I've been more easily able to take advantage of different backup software, such as Veeam's new-ish free "client" product.
TECT's hardware is very expandable. I have a single quad-core CPU and 16 gigabytes of RAM right now, but I could have a better CPU, two CPUs, and up to either 96 or 192 gigabytes of RAM. This capability, coupled with the fact that I have a pretty good disk system, means that I can easily run virtual machines on TECT.
The problem has been what to do with the existing installation. So far, I've made the first step by moving "TECT" into a Hyper-V virtual machine, running on an older business desktop, called finnmark.
The next step is to both upgrade the TECT hardware and install a newer operating system on it. It doesn't matter what order I do this in, as both need to get done before I can realistically start moving beyond the installation of SBS 2011 Standard.
The biggest problem is that I'm actively using everything that SBS 2011 Standard offers, and I would like a few more things, such as PBX functionality, and I also need a place to put a Linux or BSD shell host.
There are a few different decisions or possible moves forward. In the five years I've had TECT, smaller servers have become massively more powerful. Today, the PowerEdge T130 can run 64 gigabytes of RAM and has CPU options that are much more powerful than what I've got.
Another option is to look even smaller for new servers: If I split all of the different tasks I want to do onto separate physical boxes, I can simplify backup by using RDX docks or simple external hard disks for each system, and change the upgrade cycle for servers by keeping one on hand until the application (I'm looking at you, SharePoint) needs a lot more.
The other thing this does is lets me do is tune each system for the task it'll be doing. Because Exchange claims to need (compared to SharePoint) almost nothing to run, I can probably put it on a pretty modest system. I also know in advance that my Exchange mailboxes are relatively small, so I can build the whole system onto a single "pretty reasonable" SSD.
That said, upgrading TECT is probably still the best idea. The RAM is expensive, but it costs less than a good NUC, some RAM and a storage device or two, and that's for just one of the many needed systems. Perhaps one of the less expensive NUCs (or other low-energy mini-PCs with gigabit Ethernet and USB 3.0 ports) has a place in management or for backups.
The important decision points for TECT are how much RAM to get, when to get it, and what tasks to spin out of the existing TECT virtual machine first. Despite the fact that the Xeon E5620 in TECT is now outperformed by most mainstream desktop CPUs, it does its work well, and the real problem on newer and smaller problems (even new Skylake E3 boxes such as the Dell PowerEdge T130) is that it's impossible or inordinately expensive to get enough memory to go down a tier or two in hardware, to justify upgrading a new machine instead of upgrading TECT.
TECT upgrades in two phases. It is a two socket box, but I only have one CPU in it. Without upgrading the CPU, I can put 96 gigabytes of RAM in it. If I buy a second CPU, or two new (ostensibly faster CPUs and/or ones with more cores), then I can re-distribute the memory or buy more of it and have a total of 192 gigabytes of memory for the whole system. The next upgrade target would be the disk controller and the disks. I currently have the maximum number of the biggest disks that TECT's PERC6/i disk controller will run.
Upgrading the disks would be nice, but I think by the time I really need enough more to justify upgrading the internal disks on the server, I am also going to need or want something such as a GPU, to run more virtual machines (or do something such as run a more involved infrastructure with high availability). There are a few different options today. At the same product level, new servers have quite frankly jaw-dropping capabilities and capacities. The PowerEdge T630, for example, can accommodate eight times as much RAM (1.5TB) as the T610, and has space internally for 18 3.5-inch disks, plus amazing processors such as two of the 18-core Haswell-EP Xeon CPU.
Though, the other option is to consider a switch to the more traditional "virtualization home lab" setup and buy smaller hosts and run different VMs on each one, possibly even deploying a beefy network and a SAN on the way. That's intriguing, but I question whether it's necessary or worthwhile in my environment, where the main goal isn't necessarily to learn the virtualization technology, or to have actual high availability (although learning virtualization and a few other things has been an incidental part of the whole process.)
None of this even accounts for backups, which have become more difficult since switching to a virtualized setup. More on that later, though.
The next step is to get a new operating system installed on TECT. After that, I need to decide what to put on that system and when to upgrade it. Right now, "virtual" TECT runs on a system with the same amount of RAM, but the bare metal Hyper-V server software uses a lot less RAM than a full installation of Windows Server, so I will need to delay moving the old virtual TECT back to physical TECT. It's probably worth getting a first round of new RAM as soon as I can, so that I can move "TECT" back onto "TECT," use the TECT hardware itself for management and backup tasks, and also move one of its tasks (perhaps files, an easy enough task to move) into a separate virtual machine.
It would be interesting to use TECT as a bare-metal Hyper-V host, (ignoring the different software licensing impacts) because that simplifies (or at least sufficiently changes) some of the presumptions about how TECT must be managed and how backups can and should be made. For example, the free edition of Veeam Backup and Replication can happily back up virtual machines on a bare metal Hyper-V machine. A cheap NUC, a Windows-based NAS, or one of my old laptops can easily do this task. The potential problem, which I'll need to investigate, is that few free backup products will back up both a Windows Server machine with the Hyper-V role, and the actual Hyper-V VMs. Perhaps it's just better to think of those two things as separate entities, even though they're running on a single machine. I have time to play with that later.
For both simplicity and licensing reasons, I'll almost certainly run a full Windows Server instance directly on TECT, even if I end up with another machine to manage it. It's interesting to consider though, as this was where I was headed when I had topham, which was a cheap machine I bought for the purposes of running backups and a management console.