Enterprise-level PrimoCache and the future
Posted: Fri Dec 20, 2013 7:53 pm
Hybrid Storage is something that's currently only offered on Windows in combination with certain SSDs and with high-cost enterprise-level software. I think it'd be in PrimoCache's best interest to keep at least an unsupported version of what they have now free/nearly-free to home, non-commercial users (with some reactivation cycle, so it just can't be install-and-forget on a machine), because I'm sure you'll get far more money from making a scalable Enterprise-level solution with a per-machine/user license. Here are some things I'd like to see from a "PrimoCache Enterprise" in the future:
- L1 (RAM) and L2 (SSD) separated in options, with separate algorithms/templates, but keeping the option to consolidate their overhead.
- User choice in algorithms/templates with IOPS optimizations specific to 512K, 4K, Large Queue Depth, and Latency (and combinations of these). I would expect these to only be available/supported in the Enterprise editions.
- Maximum and minimum file size tuning for what to cache. Again, I would suspect this to be Enterprise-level.
- NTFS support with folder, attribute, and ACL whitelisting and blacklisting to allow not caching critical things, at least for deferred write (currently a feature totally unsupported by Windows's own write caching). Enterprise-level.
- Windows RAID Volumes and Storage Pools supported (right now PrimoCache treats these as separate disks). I think this would be all versions.
Remember, each user's needs may be different, and PrimoCache could solve problems with simply a different algorithm/template that would otherwise require rebuilding a server's hardware and cost thousands upon thousands of dollars. I've been experimenting PrimoCache to solve problems especially with small business servers that I'd otherwise try solving with switching to Linux w/ bcache, or switching to ZFS / virtualized ZFS (of course, I don't use PrimoCache in production, yet!).
Another reason to give out a PrimoCache Free version, would be to require users to allow automatically submitting reports on how well their cache is performing and require them to upgrade to newer builds after some amount of time (perhaps only keep Beta free, and expire old Betas after some time), thus making these Free users into guinea pigs that you can mine for precious data onto how to improve the algorithms and fix problems before they're found or implemented in your paid Enterprise-level products (as well as more feedback). I don't think providing PrimoCache in this way would hurt the brand at all, even if there's a really buggy Beta build (as long as you acknowledge this and let people download older versions). A lot of different smaller companies that offer unique Enterprise-level software have significant success with this method of a bit more limited, nagging version that is otherwise free to home users that is there to collect lots and lots of feedback and test new features, improvements, and fixes.
- L1 (RAM) and L2 (SSD) separated in options, with separate algorithms/templates, but keeping the option to consolidate their overhead.
- User choice in algorithms/templates with IOPS optimizations specific to 512K, 4K, Large Queue Depth, and Latency (and combinations of these). I would expect these to only be available/supported in the Enterprise editions.
- Maximum and minimum file size tuning for what to cache. Again, I would suspect this to be Enterprise-level.
- NTFS support with folder, attribute, and ACL whitelisting and blacklisting to allow not caching critical things, at least for deferred write (currently a feature totally unsupported by Windows's own write caching). Enterprise-level.
- Windows RAID Volumes and Storage Pools supported (right now PrimoCache treats these as separate disks). I think this would be all versions.
Remember, each user's needs may be different, and PrimoCache could solve problems with simply a different algorithm/template that would otherwise require rebuilding a server's hardware and cost thousands upon thousands of dollars. I've been experimenting PrimoCache to solve problems especially with small business servers that I'd otherwise try solving with switching to Linux w/ bcache, or switching to ZFS / virtualized ZFS (of course, I don't use PrimoCache in production, yet!).
Another reason to give out a PrimoCache Free version, would be to require users to allow automatically submitting reports on how well their cache is performing and require them to upgrade to newer builds after some amount of time (perhaps only keep Beta free, and expire old Betas after some time), thus making these Free users into guinea pigs that you can mine for precious data onto how to improve the algorithms and fix problems before they're found or implemented in your paid Enterprise-level products (as well as more feedback). I don't think providing PrimoCache in this way would hurt the brand at all, even if there's a really buggy Beta build (as long as you acknowledge this and let people download older versions). A lot of different smaller companies that offer unique Enterprise-level software have significant success with this method of a bit more limited, nagging version that is otherwise free to home users that is there to collect lots and lots of feedback and test new features, improvements, and fixes.