That the Shark 2? Or What ever they call it now?

Post #205,941 by folkert 5/4/05 6:06:10 PM Reply	That the Shark 2? Or What ever they call it now? If it is, make sure it has ZERO architecture commonalities with the Shark. All I can say, if it is a cabinet the size of 2 Industrial Coolers (typical size) and is driven by 2 AS400s inside... Get the the HELL away from it. Not that the 400s are bad... It is just the Interconnect Fabric and Framework... SUCKS The SSA setup insde of it work lickety, butyou are hampered byt LVDS (Low Voltage Differential SCSI) if you use the SCSI interfaces. Fiber channel requires a concentrator. 2 if you want redundancy... but only one can be hooked up at a time. It more like a dead-spare (not Hot-spare) Multipathing ... does, but at a severe penalty if all the pathes are not on the same controller. the path switching is a pretty good penalty. If it has a max of 8GB of memory, dump it. If it only allows 500GB per shelf toss it. Multi-Lun setups are very nice as long as the LUNs are all defined on the same controller, if not... some LUNs may never ashow up properly. Also, restrictions on the EXACT kinds of controllers that will work with the machine. We were imited to 5 different crappy SCSI cards. Fiber channel wasn't even an option until 1 year after we got ours. Then they said a fork-lift upgrade would be best... But they forgot we bought the upgrade with the machine, forcing them to swap out the AS400s for faster more capable ones and the total of 5 primary FC ports and concentrators to connect them to. Flash-backup doesn't, well it does... just that it freezes the the data on it for as long as it takes to "flash" it. Was claimed to be transparent. Multi-homing with remote siting and long Fiber Channel, does work but at a HUGE COST. The remote and local SAN must be exactly the same. It is "suggested" if you do backups with a robot/changer/multiple tape drive setup, you replicate this at the remote site as well. One last thing, if you gotta move the thing, IBM will not honor a thing if they don't move it for you. No, not moving to a new building, moving in the same room. Even though it is on nice rollers and very easy to move... Same thing with the tape Librares. Now, if this one you are looking at has none of these issues or just the IBM must move it restrictions... It probably has been re-designed to cover a few of those problems. I'd ask for a "Live HOT Demo" with it being hooked up for you Dev side of things. Pound the hell out of it. And make sure everything works as advertised. ME? I am skeptical, based on trends I am seeing, get a EMC that does SATA and lotsa SATA. IBM, HP, SUN and others are not seeing the light quickly enough. -- [link\|mailto:greg@gregfolkert.net\|greg], [link\|http://www.iwethey.org/ed_curry\|REMEMBER ED CURRY!] @ *i_wethey* [image\|http://www.danasoft.com/vipersig.jpg\|\|\|\|]
Post #205,952 by broomberg 5/4/05 7:36:38 PM Reply	Ya know, with all that ranting - you could of googled it and saw it was the FastT600 renamed, which in turn is a rebadged Metastore. I've bought 6 of them over the years. Mid /entry level, low storage, low number of FC (or SCSI) adapter. Metastore is the most common midlevel box, resold by most of the non-pure disk players. Tell me how many systems to need to speak to it, how it will be connected (FC or SCSI?),how much storage each needs, any clustering issues, and typical usage patterns. I'll spec you out the corresponding EMC gear, and then you can let them battle each other for price. Most likely cut your cost from 30-50%, no matter which one you choose.
Post #206,006 by folkert 5/4/05 10:49:58 PM Reply	You know me... Never shy about ranting. Needless to say, I have some issues with the IBM Shark. -- [link\|mailto:greg@gregfolkert.net\|greg], [link\|http://www.iwethey.org/ed_curry\|REMEMBER ED CURRY!] @ *i_wethey* [image\|http://www.danasoft.com/vipersig.jpg\|\|\|\|]
Post #206,072 by mmoffitt 5/5/05 10:28:37 AM Reply	Stab at answers. This is all a part of the Grand New Scheme that's supposed to go forward over the next 2-3 years and as such, I'm speculating about a lot of things. Tell me how many systems to need to speak to it, 3 dual Xeon Dell boxen (1 box will host DataStage, 2 for File Shares) 1 quad IBM x335 (Database host - likely DB2 UDB) 1 dual or quad IBM x445 (Business Objects or Cognos) I said above an IBM x445, but my guess is I won't get the money for that and will have to do with one of the Dell boxes The AS/400 will NOT talk to it, however. how it will be connected (FC or SCSI?), FC would be the choice. how much storage each needs, any clustering issues, and typical usage patterns. This is a read crap-shoot. We have 900 Gig of Xbase data that is going to be migrated to a real database. In addition, we have 300 Gig of DB2/400 (or DB2 for iSeries or whatever TF IBM is calling it these days) that will also be migrated to another database. I was pushing Oracle, but my boss has pretty much decided it's going to be DB2. This data is completely unstructured. For instance, not a single one of the tables in DB2/400 has a primary key, there are no constraints whatsoever, redundant data all over the place, etc. and the xbase stuff is nearly as bad. So storage requirements are difficult to ascertain with any degree of accuracy. But, we're going to have to have an ODS and a Warehouse and with our growth (we just acquired another company last week for instance) I'd guess we're looking at initial storage of 2T with growth of 2-300 Gig/year. That is a huge WAG though. I'll spec you out the corresponding EMC gear, and then you can let them battle each other for price. Most likely cut your cost from 30-50%, no matter which one you choose. Any help would be greatly appreciated. I am just a developer who has been involuntarily handed all this system stuff, you know. ;0) bcnu, Mikem Eine Leute. Eine Welt. Ein F\ufffdhrer. God Bless America.
Post #206,079 by broomberg 5/5/05 10:58:55 AM Reply	Have you been pitched a configuration with a price yet? When pitching SANS, IBM and EMC love to package a complete deal, with all the cards, software, disks, and controllers listed in a manner that you don't know what you are getting. If you ask for a different configuration (more space, more systems, more ports), they juggle the deliverable, modify the price, but don't tell you what really changed. I've seen them modify RAID level to give more space, up the price, which effectively increased the price without providing me any more hardware! This is VERY important: You MUST get line item pricing for EVERY price. You need to get individual disk costs, tray cost, controller cost, cable cost, switch cost (and make sure you understand if the switch ports are ENABLED at that cost!!!), etc. GBICS are good hidden item here, at $100 per port. Software is a huge issue. The management software varies by 99% on every discussion. You must have them put it in wiritng, and the key issue is how much TB it manages, how many LUNs it manages, and how many systems it will support. What about things like mirroring, snapshots, etc? Extra cost, but they throw them in if you fight. Als, make DAMN SURE IBM does not consider it a package disk and system deal. Everytime you modify one side down, the other will increase to match. Support is a biggie - they will price it so the 3rd or 4th year will cost more than a new system! Make them throw in a full 4 years for whatver price the pitch 2 years for. It comes from a different budget on their side, so they can get away with it. Ssme thing with software - make sure they don't give you JUST ENOUGH licenses for current GB and ports - they will claim it is workgroup pricing, the next level it enterprise, and charge you 10 times as much the second you need it. You have to expect the process to take a couple of months. You have to be able to say: Your technology is interesting, but you have no "must have" items. The other guy also has interesting tech, and his price is lower. I'd REALLY like to buy your stuff, but you have to give me a reason to - and the reason must be in measurable features and cost - not the supid relationship that they will abuse the 1st chance they get. And then say: No thanks. Again and again. There are a couple of way to play timing. #1 - MAKE SURE THEY DO NOT KNOW WHAT YOUR DEADLINE IS!!!! I've seen these guys delay again and again giving a quote, slam it to you at what they think is your last second - hoping you will have no more time to research and negotiate. So, give them a sense of urgency - saying the the decision is REAL SOON - but make sure you have real time to do it 6 more times. You should talk to EMC and Hitachi. Both have comparable gear and will probably be much lower priced. You also have to make sure your boss is very in tune. They will go over your head in a heartbeat, tell your boss you are an idiot, and should not be trusted with the project. So, how many disk, what raid level, do all systems get 2 FC channel cards for failover, etc? What about training credits for you sysadmin? And LUNS creation and port mapping can be vry complex - and they will charge you several thousand $$ to do it after the initial install. Oh, and the install is extra cost! Sure, they will bring the box in and turn it on. But after that, they want more money - and they won't tell you until you've already sent them the PO!!!! Every card in every system will cost, as will LUN creation and mapping. What about OS on the servers - make sure it is the exact support version or they get to walk away and still charge you. Have them draw a picture of what you will get. Make sure you can identify every piece, get a part number and a price. Then create a spread sheet. Start juggling quanities of disks and systems, and find out when they trigger another piece to need to be increased/ These are the hidden costs that you have to catch. Also, considered SATA vs FC drives?
Post #206,081 by broomberg 5/5/05 11:05:01 AM Reply	x445? You willing to pay 2 times the price of another 4 CPU box on the off-chance you need to scale higher? Great gear - not very price competitive. On the other hand, you can get a quad Opteron from HP at 1/2 the price, and then double the CPUs with the dual cores in a year. And it will run 20% faster to start.
Post #206,070 by mmoffitt 5/5/05 10:03:06 AM Reply	How do fail rates between SATA and SCSI compare? bcnu, Mikem Eine Leute. Eine Welt. Ein F\ufffdhrer. God Bless America.
Post #206,080 by broomberg 5/5/05 11:01:19 AM Reply	Too soon I don't trust any accelerated testing, and the real world does not have enough SATA drives in enterprise environments to really know yet. Also, you sure you don't mean FC drives? Or are you being pitched SCSI drives?
Post #206,082 by mmoffitt 5/5/05 11:12:53 AM Reply	Yes, Typo, sorry. bcnu, Mikem Eine Leute. Eine Welt. Ein F\ufffdhrer. God Bless America.
Post #206,084 by broomberg 5/5/05 11:18:52 AM Reply	Here's what I did. I had them both pitch me SATA (IBM) and IDE (EMC) system. I then realized how piss-poor the IDE would run, so I had EMC pitch me the same size in FC drives, but told them I had very little leeway now that I told my boss how much I expected to spend, based on the IDE configuration. Got an incredible price. I then told IBM that EMC's FC price was way lower then their SATA price, and much better performing to boot. Which in turn drove that price much lower. Which I then pointed out to EMC - and so on. Note: I NEVER told the other a real price - that would be breaking the agreement with the vendor. I simply said the other more more attractive.
Post #206,098 by mmoffitt 5/5/05 1:04:28 PM Reply	What'd you end up with? bcnu, Mikem Eine Leute. Eine Welt. Ein F\ufffdhrer. God Bless America.
Post #206,120 by folkert 5/5/05 3:25:15 PM Reply	I think an EMC Clairion. -- [link\|mailto:greg@gregfolkert.net\|greg], [link\|http://www.iwethey.org/ed_curry\|REMEMBER ED CURRY!] @ *i_wethey* [image\|http://www.danasoft.com/vipersig.jpg\|\|\|\|]
Post #206,124 by broomberg 5/5/05 3:39:22 PM Reply	2 of these [link\|http://www.emc.com/products/systems/clariion_cx700.jsp\|http://www.emc.com/p...lariion_cx700.jsp]

Welcome to IWETHEY!