David Gagne
Forum Replies Created
-
Why only 12 cores? To keep costs down. Do you have any idea how much those Xeon processors cost? You would be complaining about cost if they offered a 24 core version for $13k, and you’d need a noisier fan to deal with the heat. Have you ever heard an XServe fire up? Yeah that kind of noise.
Each of those processors is probably around $2,000-3000. The GPUs are probably $3k each, so a decked out Mac Pro should cost around $10k.
My bet is the low end will be single GPU, 6 cores, for probably around $3k.
-
H.264 Proxies.
-
I think FW is going to die fairly quickly now that Apple is moving on. USB 3.0 is sufficient for most.
But we NEED TBolt drive enclosures to come down in price.
-
One thing some may have missed … It appears there are only 4 memory slots.
Also, I’m not sure where they fit the second CPU? If you look at the pictures @ apple.com it looks like it only has one slot. So… Maybe a next-gen E5 with 12 cores? I believe the E5-2600 V2 will top at 12 cores.
-
In terms of resiliancy, it’s not a hard math problem.
Raid6, you have two drives that can fail without loss. Raid10 is a RAID0 (stripe) across RAID1 (mirror) sets. In this you can have 1 drive failure from any of the sets without loss, but if both drives fail in a set simultaneously, you have loss. But to quote someone I know, “I’ve never had drives fail faster than I can replace them one at a time.” If you have RAID5, and a drive fails, and you replace it same day, you will not lose data.
But lets say you are the most paranoid person on earth. You could take 16 disks and do 3 sets of RAIDZ3 and mirror them. This would give you the ability to lose 3 disks per set, or up to 2 whole sets and 2 disks on your existing set, and still rebuild. Of course you’d only get 2 disks worth of storage with this scheme, but YOU ARE LIVING SCARED. Basically as long as you had 2 good disks in the same set, your data would be secure.
Oh, but of course, you have a backup on tape, right?
-
A multi mirror raidz would be most resiliant to drive failure. Add in HA, UPS, and it’s never going down…
-
Just some corrections on Isilon –
1) you can grow to any number of nodes (not a multiple of 3), but three is the minimum per NODE TYPE. Meaning, you need a minimum of 3 NL nodes for an NL cluster. But you can add just one if you like to get a 4 node NL cluster.
2) On performance, I don’t know any specific numbers to their current offerings (mine is 4 years old now), but I know you can add performance by adding S or X Series devices, and you can also add a “performance accelerator” node.
3) The cost isn’t just the “per tb” costs, but also the per year costs. Their support costs are unbelievable. My cluster is now unsupported and ready to be put to rest as soon as I can replace it with something cheaper.(Anyone wanna buy a 20TB Isilon cluster? Didn’t think so…)
-
Yeah a multiport ethernet card would just be for adding ports for more clients (I take it you use wireless for internet and the two gig ports for afp?).
-
If you are AFP sharing over ethernet, that is NAS not SAN. You are correct that NAS does not need the extra HBAs and you have everything you need right now.
The advantage of SAN, however, is that you remove the bottleneck of your server. Each client accesses the storage as if it owns the disk, and talks to the storage directly via fiber channel. Hence the need for HBA cards and fibre channel switch. It also requires 2 metadata controllers (servers) and software. Metadata controllers manage the san but the traffic does not flow through them. XSAN is free, so all you’d need is two mac minis with thunderbolt to fibre adapters for the controllers.
If all you want is “both drives merged” but not actually gaining any performance increases, I’d probably recommend taking a look at cluStore. It combines the drive in a virtual manner to make it feel like one drive even though it’s two separate drives. It provides no real benefit besides the logical organization of your files.
-
If you’re interested in Isilon, I’d say that Coraid is also worth a look. Again it’s IT focused, but their ZX series gear is pretty slick for large media stuff.
You get dual heads for increased bandwidth and failover, ZFS filesystem, NFS mounts, and you can grow on the fly by simply adding shelves of storage or even individual drives. Also cool is that they don’t charge an arm and a leg for drives or support, so it’s really cost effective to grow.
There’s a lot of options here for growth, you can add SSDs for faster read/write cache, SAS if you need IO, or just cheap sata drives for more capacity.
Also it’s based on Solaris and so you don’t have to deal with all the proprietary commands and horrid support of EMC/Isilon.
It’s definitely not for the smaller shops, but for larger operations it might be fitting. The cost savings start to kick in at around 200TB.