6 ms·

This is not a stable success. The advantage of RHEL over other posix-y flavors is 1) cost, and 2) standardization of server build and software packaging. Solar

This is not a stable success. The advantage of RHEL over other posix-y flavors is 1) cost, and 2) standardization of server build and software packaging. Solaris / AIX / HP-UX still get plenty of mileage in the legacy application (Oracle!) space. It's going to take the enterprise longer to migrate to systems like CoreOS, but RHEL doesn't have the kind of automatic lock-in that some of these legacy systems have. As far as I can tell, RH is focusing revenue growth strategies on things like OpenStack, and similar product suites, so the company will survive. But by the time RHEL 7 comes out, what's it going to offer anyone in the datacenter? The days of the monolithic distro are numbered.

mattzito 13y ago
I'm not arguing it's a forever stable success, because the industry moves too fast for that. And yes, this is why RHT wants OpenStack and better diversification. But speaking as someone who literally based their whole business on managing Oracle on *nix, Solaris/AIX/HP-UX are dead platforms. Over 80% of the Oracle servers our users are managing with our software are Linux. Of those, probably 80% of them are RHT, and 15% are Oracle Linux. SLES and CentOS barely register. There is one more massive advantage that RHT offers vs. anyone like CoreOS - platform stability. Things like driver interfaces, library versioning, kernel versions all make it much easier for ISVs and corporate engineering teams to certify and manage RHEL releases. They're not a fast-moving target. Over time that will become less relevant, I agree, but it's been a huge factor in their success.
- dvanduzer 13y ago
  I only meant to pick on the "RHEL and Windows" remark. The OS is only a small part of the datacenter. I can't quite speak for HP-UX, but Solaris and AIX are definitely not dead platforms. My main claim is that "Big Iron" is going to outlive RedHat in the application platform category. Oracle the DB is one thing, but the sprawling ecosystem of non-database applications that Oracle has accumulated over the years aren't going to keep needing heavier distros. A lot of that stuff is moving to VM image instances anyway. What a platform like CoreOS (or other ways of combining LXC with your hardware infrastructure) offers is separation of two major categories of concern. Driver interfaces and kernel versions are hardware sensitive. Application dependencies on specific library versions far less so. Your vantage point makes sense, though. Databases are really the only application left in the enterprise that haven't seen this decoupling from hardware. But that doesn't bode well for the total number of Red Hat Linux installations.
  - mattzito 13y ago
    I realize the OS is only a small part of the datacenter, but it's most of RHT's business, so that's what I focused on. > I can't quite speak for HP-UX, but Solaris and AIX are definitely not dead platforms. My main claim is that "Big Iron" is going to outlive RedHat in the application platform category. I have a hard time seeing big iron becoming anything other than the open systems equivalent of legacy mainframe. Bear in mind that kernel interfaces aren't purely about hardware - it's about repeatable, predictable performance and behavior profiles. Watch a bunch of bank server platform engineers lose their shit when memory performance under pressure changes after a kernel patch. And remember that in the traditional enterprise, it's not just about stripping down the OS, it's about all of the other tools that get used for performance, security, backup, monitoring, remote execution, change management, privilege escalation. I actually am working on some research on things like LXC at my day job, and one of the challenges is that containers are great until you want to run a bunch of stuff inside the container. And my vantage point was at one time databases, and that's still the market segment I know the best. But I work on everything server and cloud-related, so most customers I talk to are just traditional enterprises.

mattzito 13y ago

I'm not arguing it's a forever stable success, because the industry moves too fast for that. And yes, this is why RHT wants OpenStack and better diversification. But speaking as someone who literally based their whole business on managing Oracle on *nix, Solaris/AIX/HP-UX are dead platforms. Over 80% of the Oracle servers our users are managing with our software are Linux. Of those, probably 80% of them are RHT, and 15% are Oracle Linux. SLES and CentOS barely register. There is one more massive advantage that RHT offers vs. anyone like CoreOS - platform stability. Things like driver interfaces, library versioning, kernel versions all make it much easier for ISVs and corporate engineering teams to certify and manage RHEL releases. They're not a fast-moving target. Over time that will become less relevant, I agree, but it's been a huge factor in their success.

dvanduzer 13y ago
I only meant to pick on the "RHEL and Windows" remark. The OS is only a small part of the datacenter. I can't quite speak for HP-UX, but Solaris and AIX are definitely not dead platforms. My main claim is that "Big Iron" is going to outlive RedHat in the application platform category. Oracle the DB is one thing, but the sprawling ecosystem of non-database applications that Oracle has accumulated over the years aren't going to keep needing heavier distros. A lot of that stuff is moving to VM image instances anyway. What a platform like CoreOS (or other ways of combining LXC with your hardware infrastructure) offers is separation of two major categories of concern. Driver interfaces and kernel versions are hardware sensitive. Application dependencies on specific library versions far less so. Your vantage point makes sense, though. Databases are really the only application left in the enterprise that haven't seen this decoupling from hardware. But that doesn't bode well for the total number of Red Hat Linux installations.
- mattzito 13y ago
  I realize the OS is only a small part of the datacenter, but it's most of RHT's business, so that's what I focused on. > I can't quite speak for HP-UX, but Solaris and AIX are definitely not dead platforms. My main claim is that "Big Iron" is going to outlive RedHat in the application platform category. I have a hard time seeing big iron becoming anything other than the open systems equivalent of legacy mainframe. Bear in mind that kernel interfaces aren't purely about hardware - it's about repeatable, predictable performance and behavior profiles. Watch a bunch of bank server platform engineers lose their shit when memory performance under pressure changes after a kernel patch. And remember that in the traditional enterprise, it's not just about stripping down the OS, it's about all of the other tools that get used for performance, security, backup, monitoring, remote execution, change management, privilege escalation. I actually am working on some research on things like LXC at my day job, and one of the challenges is that containers are great until you want to run a bunch of stuff inside the container. And my vantage point was at one time databases, and that's still the market segment I know the best. But I work on everything server and cloud-related, so most customers I talk to are just traditional enterprises.