Actually that was a comparison that started as a challenge by Red Hat bloviating all over the place. FINALLY, Microsoft took them up on the challenge and slaughtered Red Hat. For damage control Red Hat pooh-poohed the results when THEY are the ones that pretty much set the table how things would be done.
HOWEVER, You missed a key point, the gr&d. That was then. We started out on shared hosting but things grew pretty fast. They promise unlimited and call it abuse as soon as you lay a load on them. For dedicated, people were selling Linux hosts. The Linux kernel was a piece of work for a long time which is how I ended up on FreeBSD. Linux could NOT take a load without falling over. It can't be getting knocked over that easy today or people wouldn't use it on shared hosting, and back then none of the big hosing sites like ThePlanet, iPower, etc. didn't use it for shared hosting. Linux was the geeky thing to do so there were hosting providers who had and and said they could do the job, no problem. No they couldn't. One said they would set up FreeBSD machine for us and as soon as they did, they said we had a week to leave because of the traffic, so unlimited gets called abuse, and that is how they get out of that. Server fees and Google ads were about even. When I talked to pinciple at ThePlanet he said we were too small from a server volume standpoint and way too much traffic for most vendor's infrastructures. He said then, keep in mind these three things. For shared hosting or high traffic hosting, you want FreeBSD, for applications you want Linux, for Microsoft apps you go Windows. SAVVIS took us in. (They were/are? considered a national asset and guarded by the military and where most of the world's banking servers were at that time.) I talked to real engineer there and he gave me the same line. He said, you want FreeBSD, and if you need a control panel you want DirectAdmin. I'd only ever heard of cPanel. He said cPanel takes resources you can't afford. I checked it out and it was in beta. I mentioned that. He said, I'm not going to recommend you use it in a production environment. Well maybe he wouldn't but I was kind of stuck. Instantly, we could handle the traffic reliably and making money. It would be a long time before Linux could stand up. Linux would get to load 8 and things fall over. FreeBSD could be under attack at 400 and you could log in...slowly, and defeat the attacks. Heck, email wouldn't stop until 65-79. They contacted us a couple times and said they would be will to move us in their data center for better response if we wanted. We took them up on it. There was one part that was a little frustrating, and that was response time sometimes, because it was supported. It might take and hour or more before someone could get to it and you were never sure how things were set up. Then I learned of how a group left ThePlanet and started SoftLayer. I didn't think much of it at the time, but after they were running for about 5 months, I decided to give them a call. WOW! They had everything I needed with a VPN, portable IPS, KVM, virtual power switches, lots of bandwidth, and good prices, and bare metal where I could build what I needed, and a lock-box where I could copy up an ISO to boot from. SAVVIS had been VERY good to us and gave us incredibly powerful servers, but Softlayer is where I needed to go, and I could get ThePlanet-type infrastructure to work in. I decided I always wanted to try virtual, and had been dabbling with VMWare before they were VMWare and in a basement corner at COMDEX Chicago and the hardware wasn't there and the software cost more than the extra hardware. So, I got ESX and put it on a new server before we moved out of SAVVIS. VMWare engineers said that we were no match for the purpose of the server. My argument was, your sales pitch is to consolidate underutilized servers which means adding load. They said it would not work very good. My partner said we could add the smaller sites and see how much it can take. I thought about and decided, no, I'm going to install the OS bare metal and run the stars out of it and measure the results. Then I'm going to put ESX down and the OS on top and run the stars out of it and measure the results. I found a guy who had been running it for a while and an engineer at VMWare. I learned a lot about how my thinking had to change between the oh-so-logical hardware and ESX and VMs. However, they learned a lot from me too. They learned that for disk, VMFS doubled my disk performance over UFS on bare metal. On the CPU, when you are set up right, you will lose 25%., but cores were comparatively cheap even then, and ESX didn't take much RAM itself. Now throw in over provisioning of all resources, snapshots where you can roll backward and forward to any point in the tree, backups you can mount, I never wanted to see bare metal again. Windows was trying to get into the web game and for a short time, I could get 2003 Server Standard 32 or 64 for free, and the 32 came with the ability to do NFS. So, I ordered two more machines with 2003 server, did P2V on them, hosed the servers and put the newly releases ESXi on them, with the 2003 Server Standard VMs to project NFS volumes. The deal with ESXi then was it was very expensive if you wanted to be able to back it up, BUT it could also store VMs on NFS volumes. Bingo! I had a free virtual infrastructure and automatic backups between servers at night using the standard tools that come with ESXi. Everyone told me that the backup performance would be poor and I'd be better with a bare metal target or a ?NIX VM NAS in FreeBSD or Linux. I found that to not be true and Windows was far more stable. I opened up ESXi itself to run scripts that would be run by Windows programs that come with PuTTY and a Windows graphical scheduler to manage the backups, check for updates required on the ?NIX and Windows VMs and E-Mail me the results every morning 15 minutes before the workday starts. I don't have a dime in operating systems. A guy spent a lot of time later make a program to called GhettoVCB to do what I was doing and I went with it because he keeps up the updates to this day. Those 2003 Servers are running fine to this day as well. As things progressed, FreeBSD EOLs got shorter and shorter, an then you get problems with ports, and then it doesn't make sense for DA to put a lot of work into it, and now after 12, not at all, and how I got back to Linux, but it might be a horse we can ride, and if not, we really don't need a control panel.
I don't know if you've used ZABBIX or not, but it is tremendous for analyzing because it keeps history and graphs over time that are highly detailed. and you can see in a snapshot in time what was going on when and merge graphs to show say MySQL loads, processes, Apache, CPU, network, traffic, memory, and disk traffic. It will send texts, e-mails at certain points if you wish and the reason it beat the stars out of nagios was it is all in one integrated package. It's also easier to set up your own Being able to see trends is huge too and then be able to investigate what was happening at certain times and have available everything that was going on at that time.
Windows has one HUGE thing going for it, and that is its API. That caused the IBM/Microsoft split. IBM took the OS/2 that was ready to go, and Gates took Portable OS/2 Lan Manager which wasn't ready. and why when they split, IBM thought they were going to slide OS/2 out front with its ability to run Windows programs. The split was over Gate's vision for the Windows API and IBM's vision. Not only did the IBM OS/2 API vision not well thought out, the quality wasn't there. A friend who owned a large software language company at the time was pushing Intel OS/2 and I was one who played with both and proved to him that the betas of Portable OS/2, renamed Windows NT, were more stable. He blamed his developers for problems he was having. One day he called me and said, You were right. The OS/2 API is not just inferior to get where you want to go, it is impossible because it is loaded with bugs. To this day, there is no match for Win32/64. Not even the MAC comes close, and certainly not ?NIX. Microsoft had the business mobile world tied up until Balmer flushed it and went with a rinky-dink Zune Gaming OS that he renamed Windows Phone, closed off multitasking to Microsoft-only, started their app store, and decided he was going to be Apple. The iPhone had few apps, and Android wasn't something people had heard of. However, every phone manufacturer outside of Apple needed an OS, Microsoft was putting a lock on them with no apps and the Windows phone developers were hamstrung, so they all left for Android, which was more open like the previous Windows environment. That was an incredible gift to the ?NIX world. That rinky-dink OS would later be known as Windows RT on the tablets and nobody wanted that either. Windows couldn't sell a tablet until they put real Windows on their tablets with the WIN32/64 API on it that could run real Windows programs instead of the zune/metro/modern/universal apps. Win32/64 was been tweaked and refined with assembler in areas and it has no match for efficiency at what it does nor the primitives needed by developers, sofves DLL Hell, 10 years of updates, granular security, ties between client and server for apps and licensing, etc. HOWEVER, these days, Linux can beat the performance of Windows in a lot of areas because it doesn't have any this going on. However, for technical people running a web server, they don't need all that. Microsoft was granted the rights to (from Sybase where MSSQL came from) to sell to the ?NIX world. Previously, they could only sell it to the Windows world. Nothing comes close to beating MSSQL on a normalized database, but even THEIR benchmarks show that Linux can run it faster than Windows at very high performance levels.