btb4

Member
  • Content Count

    282
  • Joined

  • Last visited

Posts posted by btb4


  1. At the moment, most RAID assemblers are using RAID 5 for professional video editing systems. For example, Rorke standardizes on RAID 5 for their 844/x (10 bit uncompressed, up to 8 simultaneous streams) certified arrays. I also know that Rorke has gone to SATA for their Galaxy series, and either way (PATA or SATA) they pretty much stick with Hitachi 7K250 series drives. Take that for what you will, but Rorke/Bell is probably one of the best, most highly regarded assemblers for professional video arrays so personally unless I have a good reason to do otherwise, I tend to do as they do (which includes buying their products in some cases).

    Depending on the system you are using and the sort of editing, I do not necessarily think that a 3 drive array will be either fast enough for professional video, nor enough storage. Even if it will work for acquisition, you may not be happy with the performance during post. With 1.7T on line, storage is still always a problem for our 844/x.

    From thos that you listed, just based on overall assesment of the company itself, I would go with LSI first, 3ware second.


  2. Have used it on an Adptec 3400 and on two Mylex AcceleRaid cards. In the case of the Adaptec it was simple expansion - added a comparable number of drives to the 2 remaining channels as were on the initial two in the original set. In the case of the Mylexs it was a reconfiguration.

    I can not say this would be "often" - and in the case of the Mylex expansions it was a matter of convenience, but in the case of the Adaptec that had as I recall 2 channels of 8 36G drives to make that a 32 drive set while the system was operational was pretty much a lifesaver.


  3. It is also probably worth mentioning that some OEM products are not the same as their retail namesakes, and indeed in some cases they are merely licensed to a particular assembler for manufacture at their own/hired facilities. Also, some drivers/OEM products are hobbled so that you may not use the generic drivers provided. Probably NVIDIA & their universal driver program had a lot to do with reversing this trend (7 or 8 years ago, for example, you could not use drivers from the chip manufacturer on most boards built by others).

    Boils down to cost, as others have mentioned, as even every time someone DLs a driver from your web site there is a real cost that aggregates pretty quickly to serious $.


  4. Comparing the "ouside" - visible portion of a WD800JD and the Raptor the boards appear fairly simular - up to the bridge. I'll probably - um, that is hypothetically - remove some boards and see what the undersides show. Think the boards can be swapped? Wonder what a WD740JD 10K would do....


  5. I just wanted to thank everyone for their input and encourage more users to give their 2 cents on what they think I should do with my companies servers!  B)

    Well, something that I did not see anyone reply to is the notion of "DVD burner for daily backups". The standard for backing up systems such as you describe is tape - pretty much period, other than "what kind?" (DLT, AIT, etc.). This should be running continually - doing potentially multiple incrementals an hour, with potentially full backups according to whatever schedule floats your boat.*

    Your general approach is contrary to the overall trend, and probably a not so great business model. For a whole bunch of reasons, most folks (as has been suggested by others) are going for the "many mediocre systems" approach compared to the "few muscular systems" method. Take just one of your apps in consideration: DNS. You have suggested building only one DNS server, however you need at least two. Three or four are better. You can build a basic server-class box perfect for DNS use for less than $1K - and that is still very "server class" (Intel or equiv server mobo, certified 1U case, etc.). The great thing here is that those boxes will be 100% dedicated to DNS, and because they are so cheap it will never even cross your mind to use them for something else "just a little bit". Overall you will probably be better served by this approach.

    Your optimistic scenario about the number of users etc. will allow you to continually add boxes of the same configuration to add capacity as required, in easy and inexpensive steps, and as you have the capital. Make no mistake about it - this is in no way a 5 year (or even 3 year, or even 3 day) "fire and forget" setup that would will be able to build no matter which approach you choose. You are talking about a very busy occupation here for somebody.

    *Oh, and one of those <$1K boxes would make a good dedicated backup box w/tape, too....


  6. I would NOT, however, make/break a P68 cable connection on a system performing IO on that same bus.  Unless you short something out by misaligning the connector, you're unlikely to do any "hard" damage, but data loss or a system crash is a definite possibility.

    For this reason I would never do it, period. Yes, you can get away with it, esp. if there is no OS drive or whatever on the channel, but if you plug&unplug SCSI often (sometimes 100s of times in a week) and you are not in Centronics world sooner or later you will mis-seat a cable. This will nearly always ruin the cable, may ruin the connector on the SCSI device, and if powered can cause rapid nastiness. OTOH, in a well equipped shop if the power is OFF this is not a huge deal - toss the cable in the NFG cable pile, scowl at going through $1000s in cables every year (SCSI, VGA, RGBHS, RGB, XLR, etc.), check the connector, be more careful....

    In fact, it is sometimes possible to mis-align a 68 pin SCSI cable and not realize it until powered up, which is why I always double-check.

    A tape drive on its own card may be an exception to this, but why chance it?

    As to the screws, my eyes are too aged these days to be able to always tell, so I always line up some "usual suspects" and lightly test by hand straight into the drive. If the screw goes in easily like that, then I know I am good. OTOH, on one or two occasions I have deliberately used screws that were too big due to a need to hang a drive in a particular, unusual way (kiosk or whatever). Big screws sometimes seem like the only way to really mash a drive in, though this is probably not recommended for long-term use.


  7. btb4,  any particular button models you've worked with?

    I'll try to remember to check next time I am in the shop. I don't think they are anything exotic, though - for stuff like this we try to find things that can be easily replaced on the road if needed (CompUSA, BestBuy, whatever).


  8. What do you think of this.  An LSI Logic LSI-21040 with a 68 to  50 pin adapter.  The controller will detect that the Jaz drive is a legacy standard. I will still have an available U160 channel for getting scsi drive if a real bargain comes along.

    I'd try it first. It will probably work, though you may need an in-line terminator. Jaz drives can be fussy. Understand that any channel where you connect a Jaz will slow down to Jaz-compliant standard. It has been too long since I've used a Jaz to recall what that is. Unless you are short slots, you are probably better off getting a cheap adapter to run your Jaz and save the U160/320 stuff for when/if you get a drive. As long as you stick with the same manufacturer you are probably OK.


  9. Usually not. I hate absolutes like "of no use" - it really is a matter of opinion, based on the practical knowledge that for very few apps will there be any meaningful benefit - for some, none at all - coupled with the very many disadvantages.

    You usually see RAID0 in two places: "Performance" PC assemblers, who know that even if there is little real-world benefit to RAID 0, it will help get there systems to the top of the benchmarks. Low-cost RAID manufacturers of specialized RAIDs, like Medea will also use RAID 0 in their more cost-conscious model lines.


  10. So if I wanted to encode an AVI file to a SVCD format it'd make no difference what so ever in Raid 0?

    That's the general notion. In all probability that sort of transcoding task will be CPU/memory limited, rather than disk limited. if your software is very kludgey the disk may make a very small difference, but usually the read/write aspect of that sort of task is less than 5% of the total operation.


  11. OK flick, could have used a bit more action - maybe some sword fights or something? Based on the hype I was a bit disappointed, esp. about the supposed "realism". I was hoping for more pestilential boil-ridden iron-age squalor.

    Since it was so Hollywood, I think it would have been much better if a few of the director's strengths had been better exercised.

    Imagine:

    <Jesus on the cross still on the ground awaiting being hoisted into position. A black-leather clad Pilot approaches and tosses a small metallic object to the supine Son of God. Pilot speaks:>

    Pilot: A man in good general health lasts two to three days before asphyxiating on the cross. The ropes that bind you are Centurion-grade high-cellulose military hemp. Due to the restriction of your dexterity from the nails in your hands, I estimate that with this sharpened bit of bronze it will take you three to four days to cut through the ropes. Cut off your hand at the wrist in maybe two....

    Hmm, yeah, OK, once the tendons go, how could he keep cutting? Still, not sure if functional anatomy was an ancient Roman strength.


  12. FWIW, Promise tech support came back and said it probably would not work. I give them full marks for both honesty and speed (I sent the email question at about the same time I started this thread).

    The reply also said something else interesting, though - he indicated that none of their boards were 64-bit. Maybe my eyes are getting tired, but they looked like they had the extra pins in the photos I've seen. Perhaps it was another card? For full-SATA and 4 drives don't the cards need to be 64-bit?