Alex Gerulaitis
Forum Replies Created
-
[Bob Zelin] “Putting together a simple system that works with AFP is a wiser choice for him”
Second that. One of my clients, who bought several DAS boxes of various sizes, does exactly that for shared storage:
– Mac Pro (or even old xServe) – $1K-3K
– RAID controller $800
– DAS 24-64TB with speeds ~ 600MB-1.2GB/s – $3-8K
– Optional: Small-Tree 10GbE cards, Netgear 10GbE switch ~ $4-5K for three systemsThe above DAS shared over AFP via Mac Pro is a fairly robust, expandable, customizable, scalable shared storage solution that can be put together using an older Mac Pro for under $5K total.
Then of course you’d have to muck with it to make it work or hire Bob Zelin for the purpose. 🙂
-
[doug nash] “I suppose, one way to manage both the data and the risk, would be to scale back on the actual size of the centralized RAID (assuming I use such) so that it only holds immediate, current work. Then just off-load finished pieces onto other media, and store those off-site.”
That’s exactly how the big boys do it: tiered approach. Back up or replicate the most critical data: source files, projects. Anything that’s temporary or that can be re-created w/o too much pain doesn’t have to be backed up.
In many cases the main NAS like SNS Evo or Small-Tree GraniteStor can be the central storage, while replicated or partially or fully backed up to a much cheaper DAS or some sort of a Drobo-like box.
[doug nash] “what are you thoughts on DLT for just that kind of thing?”
Like Bob said, LTO is the modern equivalent of DLT, and actually not all that slow at around 140MB/s uncompressed – faster than GbE speeds, for that matter. No robust long-term enterprise-class backup happens without tapes so it’s a perfect backup solution if there’s budget for it.
Tolis Group makes awesome LTO-based backup appliances for Macs.
-
[doug nash] “1) Isn’t part if the entire value of a RAID, the redundancy? Doesn’t that minimize or entirely mitigate the potential critical failure of any one drive in the group?”
It does, especially with the so called dual parity RAID schemes that protect the data against failure of two drives, and mitigate certain other issues. RAID6 is the most popular flavor of dual parity RAIDs, and everybody’s favorite with few exceptions.
RAID doesn’t protect against user error (accidental deletion), catastrophic losses (fire, theft, etc.), certain instances of data corruption, certain other hardware failures. Because the data is centralized, risks of losing all of it are higher than with individual scattered drives.
[doug nash] “2) In simple terms, how would I network all three (or four) Mac Pros to the central drive? Would I just use their built-in Ethernet ports, visa-vi a simple Ethernet router with the NAS hooked to that? Or is that way too simple to be true?”
Hardware-wise – that’s it unless you need 10GbE speeds which would require 10GbE cards installed into Mac Pros. You’d also need to configure permissions on the NAS so that client computers could access it – usually a straightforward process.
-
[doug nash] “So, I’m wondering what might be the best idea for mass storage? It would seem that a big shared drive (RAID?) would be best. But is that right?”
NAS seems to be the perfect match for your purposes, especially that there don’t seem to be any exotic performance requirements such as “three streams of uncompressed 12-bit RGBA 4K to each client”.
As always with putting all eggs in one basket, backup becomes even more important with shared / centralized storage vs. individual drives. Ease of management, monitoring, service-ability, support are up there too: if the NAS needs attention (degraded RAID, components that failed or are about to fail, etc.), you’d want to know about it right away, and have the means to fix it.
Among companies making fantastic NAS boxes are Small-Tree and SNS (both active on Cow), I hope they will chip in.
[doug nash] ” I don’t know if I need the same massive throughput as an editor might”
Decent GbE NAS boxes are pretty close to individual drives in speed, so there should be no surprises. For a little extra $$, you could go even faster with 10GbE connections.
— Alex Gerulaitis | Systems Engineer | DV411 – Los Angeles, CA
-
[Vik Narayan] “I would like to improve the realtime playback of the system and am looking adding an internal SSD into the mix by replacing the DVD drive I rarely use.”
Vic,
The general suggestion is to use the faster drive for caching, but it’s a general one – and may be useless in your specific circumstances.
When you say, “improve the realtime playback of the system”, what do you mean? Is the playback choppy? Pauses too long after hitting “play”? Scrubbing too jerky?
— Alex Gerulaitis | Systems Engineer | DV411 – Los Angeles, CA
-
It sounds like you’ve decided to use those four drives individually, but then asking if that’s the best choice vs. RAID0?
RAID is nearly always better for performance, capacity, management, usability – but comes with some risks, especially 0. Is RAID5 an option for you?
— Alex Gerulaitis | Systems Engineer | DV411 – Los Angeles, CA
-
[Kenn Schulz] “0W PSU from HP did not work (should’ve). MOBO gives 4-beep codes.”
Possibly a bad power supply? What does the manual say about 4 beeps?
[Kenn Schulz] “I want to go to (at least) a 2GB GPU and HP compatible PSU (550W?).
Any suggestions?”
Non-HP PSU will not work in HP workstations, so the choice is limited to those two.
K4000 will work with the 475W PSU, perhaps even a K5000. A $250 150W GTX-760 will work as well – probably the best choice.
— Alex Gerulaitis | Systems Engineer | DV411 – Los Angeles, CA
-
Don’t have the answer, just a question… 🙂
“tar” seems to be part of Max OSX CLI (Command language interface), and there are some offsprings with GUI not requiring UNIX CLI knowledge. Have you looked at one of those? (Just curious.)
-
[Scott Smith] “they sometimes lapse into a lingo I cannot quite follow at the speed at which they sling it… would you trust an IT solution from an IT department that realistically never deals with Broadcast type needs?”
I think you answered that question already: no. To confirm that, sling your own lingo at them at even faster speeds and see if they can follow. Ask them to do a brief proposal that guarantees compatibility with your creative apps (Adobe, etc.), gives a guaranteed aggregate and per-client bandwidth and IOPS. If they blink, start smirking or say something like, “well, your apps can’t be anything as IOPS and uptime hungry as that Oracle database we’re running”, then it’s confirmed: they can’t be trusted (yet). You can tell them to stop smirking and actually prove it. Put the ball in their court, make them do a little research on your apps – that would go a long way for a healthy relationship with your IT folks – after all, it helps to have them on your side. After doing that research, they may end up recommending the same equipment you’ve already looked at.
The other side of this conundrum is some homework of your own: preparing a decent RFP. Figure out your current and estimate your future needs in performance (bandwidth, IOPS), scalability and expandability, uptime. It can be a lot of work, but it’s worth it.
Alex Gerulaitis
Systems Engineer
DV411 – Los Angeles, CA -
Lunch would be awesome! I’ll call you Chris.