Newsroom Update
Beginning in May, a special Today at Apple series titled “Made for Business” will offer small business owners and entrepreneurs free opportunities to learn how Apple products and services can support their growth and success. Learn more >
Beginning in May, a special Today at Apple series titled “Made for Business” will offer small business owners and entrepreneurs free opportunities to learn how Apple products and services can support their growth and success. Learn more >
The Server version of Mountain Lion, OS X Server, emphasises on Xsan, but my question is what to expect, what changes will I have to make to my current configuration (4 mac mini servers + an iomega NAS) to build myself an "enterprise class" mail system and is that worth the costs or is this just good old Apple marketing?
OS X Server, 8GB, 2,93 GHz Intel Core 2 Duo
What XSan version is in Mountain Lion? I'm still not seeing an XSan User Guide newer than v2.3 which came out with Lion. Funny enough on http://www.apple.com/xsan/ the photo is a box for XSan 2.2.
Umm, I personally find it a bit odd that XSan/Stornext is referred to as a clustered filesystem rather than a shared disk file system. If you're looking for high-availability storage, then you need a distribute or distribute+replicate so that the storage itself has redundancy (not just one disk fails but something takes out the entire array). In the case of a Mail server cluster, it's the Mail servers that are redundant not the storage.
If it were me, I'd change as little as necessary, and migrate to something else entirely down the road. This gives you some time to see what, if anything, Apple is going to do with XSan before you invest more in it. And time to research alternatives. There's a lot going on in open source in this area now that do not require FC, dedicated metadata servers, and are highly scalable.
What XSan version is in Mountain Lion? I'm still not seeing an XSan User Guide newer than v2.3 which came out with Lion. Funny enough on http://www.apple.com/xsan/ the photo is a box for XSan 2.2.
Umm, I personally find it a bit odd that XSan/Stornext is referred to as a clustered filesystem rather than a shared disk file system. If you're looking for high-availability storage, then you need a distribute or distribute+replicate so that the storage itself has redundancy (not just one disk fails but something takes out the entire array). In the case of a Mail server cluster, it's the Mail servers that are redundant not the storage.
If it were me, I'd change as little as necessary, and migrate to something else entirely down the road. This gives you some time to see what, if anything, Apple is going to do with XSan before you invest more in it. And time to research alternatives. There's a lot going on in open source in this area now that do not require FC, dedicated metadata servers, and are highly scalable.
Thanks Christopher,
I was thinking about waiting too, since the conventional solutions are not as fast as Thunderbolt and the highly marketed Promise vTrak is a bit old in my opinion.
Not only the speed, but the costs too are currently not feasable.
Yeah the FC cost makes 10GigE look quite feasible. The cabling isn't exotic, and you don't need a dedicated metadata network.
I'd look at Open Connect and Open Compute, for hardware ideas. If you just want really fast reliable single pile of storage, FreeBSD/OpenIndiana and use ZFS. Multiple ways to connect. And Linux with GlusterFS over Infiniband would even get you into the 2x performance from 10GigE and less than 1/2 the price if you weren't married to an Apple mail server solution. The Mac clients could connect with GigE over NFS to that same GlusterFS storage. And of course that aggregated storage can be divvied up into logical volumes for various groups/services just like XSan lets you do.
And re: Thunderbolt, well while it's fast, it's just a bus. You still have the cost of adapting it for a particular purpose. The cost of integrated adaptation, inside a storage enclosure for example, doesn't seem that unreasonable to me. But Thunderbolt to FC for minis is $800. And if you want 10GigE, I'm not even sure there is such a thing yet, although OWC has a ~$400 PCIe enclosure, and you can put a 10GigE card in that.
For enterprise storage, Mac OS relaly isn't the platform: Not from the hardware perspective. And not from an operating system perspective [1]. And not the file system either, JHFS+/X just don't count anymore than NTFS does. There's way way more interesting stuff happening elsewhere.
[1] Windows= 5 years features updates, 10 years bug/security updates. Red Hat Enterprise Linux (or CentOS) 10 years standard, up to 3 additional years extended support for RHEL. Mac OS X is in the realm of 2-3 years, and in fact it appears anyone who wanted to upgrade servers to Lion, can't now. It barely gets stable enough before it's yanked from the App Store in favor of the latest version.
Xsan mail server clustering in OS X Server (10.8)
