80 posts • joined Wednesday 10th November 2010 04:09 GMT
We have a ton of oracle zfs storage appliances here.
1. Usable vs Raw is horrendous
2. If you utilize more than 80% of the box the performance goes downhill
3. Due to it being zfs/software raid adding additional shelves requires you to add even numbers
4. They produce hilarious amounts of heat
One of the biggest issues with Brocade is when I have device/port issues I first see an error in the main error report, and then have to visit the fabric log to start figuring out what is actually going on. The Fabric log is incredibly unintuitive and frequently useless.
So you have to upload support saves and get told "slow draining devices"..... but they can't point them out. Just they exist.
So now you come to the situation where you're trying to find a root cause and all Brocade can do is vaguely point out you have something somewhere.
So the SAN gets a bad name and management gets sold on FCoE.
"The last claim; come on. Fusion-io, LSI, Micron, OCZ, Samsung, Plextor, you name it, can all make the same claim. This indicates a weak hand if the argument has to be bulked out with thin stuff like this"
Hold on a second. Of that list Plextor isn't OEM'd by anyone (Dell maybe). OCZ is oem'd by HP and their recent spate of stupid and firmware has left most folks running. Samsung has several great OEM contracts so absolutely. LSI/Micron and Fusion-io are all mostly in the server accellerator market, so really saying they are in enterprise storage gear is a touch misleading. Yes, every tom dick and harry these days sticks a fusion-io card in a server and calls it storage, but when your company has yet to have a drive failure you're not enterprise.
Sure STEC has squandered their lead, but if they can get 2TB flash to multiple OEM "soon" then they have certainly redeemed themselves as most enterprise storage arrays are stuck in the 200/400GB range.
Re: I didn't know Oracle made storage
He doesn't. He makes servers with flash pci cards in them, just like every other lame startup with no engineering chops but the ability to buy fusion io cards. Well, to be fair, they have a lot of tape units.
I realize by talking bad about Oracle I'll get a few thumbs down, but there's a few facts about Oracle "Storage" that make it awful.
1. ZFS and software raid. The amount of usable you will get is greatly reduced due to overhead, lack of proactive sparing, and other features that require a lot.
2. Growing trays require you to grow two at once
3. Once you go over 80% utilization the head changes its performance algortyhm to greatly reduce performance
4. No thought has been put into heating/cooling, so you're adding in a lot of furnaces to your data center.
5. Prices that make EMC look competative, especially when they hit you with the software maintenance
6. It's the same terrible SUN and Oracle support folded together.
There's a reason most SUN engineer based startups went titsup (3PAR a famous counter example), they have the software chops to get throughput but almost no basis in intelligent hardware design.
HDS has a scaled down version of the VSP, the Unified Storage/VM. IT allows HDS to have the entire range of products.
Anyone saying the VMAX is an archaic platform is stuck in the DMX realm. And the Symmetrix folks using weird acronyms and weird terms are products of the 80s and 90s. The VMAX of today is fairly close to any other storage system in terms of how storage is presented, carved, etc... It's a very different animal and is only going to get more robust and interesting.
The VNX team is a fantastic group of engineers, who have benefited from being the second child at EMC. While the Symmetrix team have been aloof and "all mighty", the VNX team has been able to use all of the products, so you see better cloud integration, better recoverpoint integration, much better VMWare integration etc... EMC World is going to bring over a million IOP VNX and some other features. Neither product is going anywhere however, as long as the VMAX is the product of choice when you need a lot of back end throughput, which in the end, very few companies can match at the scale of VMAX.
As someone who has worked with multiple storage monitoring/management tools I disagree completely with you hoping other vendors follow NetApp's example with API, because they all suck. PERIOD. Always have, always will. Nobody wants to play well with others, everyone wants their own tool to be the best, and nobody wants agents. Every single time the other vendor releases new code your product from the other vendor will cease correct reporting.
So, you'll continue to have NetApp's issues with bad reporting, bad data, and a portfolio thrown together because they want to fill checkboxes. Netapp's product built on a platform designed to fill in the gaps in EMC's ECC so it's better with EMC. EMCs frankenstein of Prosphere and 3 other products they bought, Aptare's inability to stop tripping over themselves in their database, Symantec's joke, HP's disaster, and whichever lame product HDS is reselling.
There's just no money in it, so get used to hating it all.
Actually, hitting the 5 9's for "reliability" and "downtime" with any decent piece of kit is pretty easy, and the vast majority of vendors when it comes to hardware failures these days far exceed that.
But, and it's a big but, you then get into five very common situations
1. You were one of the first to use those shiny new software features that had a bug that corrupted everything
2. You are dumb enough to take that system running fine for a few years, turn it off, and move it across floor to shove into different corner
3. The third party company you subbed the upgrade to didn't do something right during the hardware upgrade
4. You didn't replace that hardware item a few weeks ago when it first failed
5. Sooner or later, you get to a situation, where due to economy of scale, the guy on the phone from support has to tell the guy on the other end to replace something important and <human> happens
So you really need three things:
1. The actual hardware reliability
2. Average amount of time after major software release final HOLY CRAP bugs get worked out
3. An idea of how much your refusal to allow remote support is going to hurt you when you release the customer support engineer licks rocks.
The worst part about the original Challenger disaster is that under the same conditions during early launch testing of the program they noticed the same exact issues with the cracking of the O rings at low temperatures.
While many bemoan the privatization of the US Space program, NASA has shown repeatedly an inability to properly respond to issues with this program and many of their nonmanned rocket programs. Put simply, they're incompetant.
Re: I'm floored
I once saw a 3bay EMC DMX Symmetrix almost fall over onto a moving guy when the floor tile buckled. Now that "monolithic" is a bunch of shelves it's a lot less of an issue.
STEC seems to be a huge liability/lawuit magnet and OCZ's reputation for drives that brick is so widespread that only HP has OEM'd them. When your only major win is the disaster known as HP you're screwed. There's a reason Seagate ended up passing on them.
"I love it when storage vendors invent new market segments: "Entry-Level Enterprise Storage Arrays" appears to be the latest one, from the brilliant marketing team at EMC. And it is always a "new" space that only the company occupies."
FROM THAT BLOG:
"EMC has its very successful VMAX 10K. Hitachi has recently offered the HUS VM. And HP has invested heavily in enhancing 3PAR in this segment."
But really, what is the point of this? Is there something coming where you talk about how enough features have moved from Enterprise level storage to Mid tier storage that the highest level isn't necessary any more? Or perhaps go more in depth about how the way applications like Oracle and MS Exchange interact with I/O causes inflated backend storage throughput that wouldn't be necessary if they just did x or y?
Are you going to steadfastly refuse to ever use the term "Big Data" because marketing at EMC thought that one up (so they tell us)? How about an amusing story about how the one time EMC Marketing tried to rename Legato's clustering software Full Time Auto Start (ahhh FAT ASS)
But just to write a column saying the term "enterprise" is used to much and then say "too expensive" kind of lends the mental image of a guy sitting on his porch shaking his cane at passerbys.
The VMAX 10K doesns't sit above the VMAXe, it is the VMAXe. When EMC did a enginuuity refresh with the 76 code, because they made it completely backward compatible with older VMAX, instead of calling it a VMAX2 they rebranded the old VMAXe the VMAX 10k and the existing VMAX became the VMAX 20K.
So entry level 10K
Old VMAX is now 20K
New "hotness" 40K
The 40K have the processors with additional cores (dedicated to back end) and you cannot get the 2.5 inch drive cabinets with the 20k.
Re: Sounds like the return of Invista
Invista lives on as part of Vplex, if you use EMC Powerpath it identifies VPLEX luns as invista. EMC combined the invista technology with Yada Yada ip to add Active-Active clustering and made it a much better product in terms of performance and scalability with much more serious array hardware.
Cisco is downright pissed at EMC and VMWare for the Nicira purchase, enought that they are thinking of ending the EMC relationship and leverage UCS' sales with another storage vendor and virtualization platform.
Citrix and NetApp bring replacements for both those. It's not a great replacement at either level, and Cisco has to be wary of what happened with both Dell and Data Domain with regards to EMC.
Citrix has great penetration into VDI. At that level storage is interchangable. NetApp has a wide range of platforms, but Cisco has to be aware of the issues with NetApp engineering being slow on software and useless on hardware.
A costly set of companies is not financially smart for a company finally starting to show progress after the disaster of home based products and failure to innovate on the networking side more than competitors. The sale of linksys will bring some cash, but Cisco still needs to revitalize it's networking component which means money.
But the bottom line is, all of the companies seem to be preparing for the end of the EMC Cisco relationship, and with one of the two rock solid with the other on shaky grounds it doesn't bode well for Cisco to end the relationship. Klayko over at Brocade would welcome the chance to step down with a resurgance of orders from EMC though.
You forgot to mention the article is a French Newspaper that has had translation done on the interview
If you read the comments in Chuck Hollis' blog, it gets more entertaining.
1. The CEO Axxana says he had the same experience
2. The writer of the article can't figure out how to post comments to an online blog
3. The editor of the article can't help but douche up the whole thing even more by telling Chuck Hollis he should have e-mailed instead of blogged when the article didn't bother to confirm in the first place his translation was correct
For those who don't know, the noted blogger Chuck Hollis refers to is Robin Harris, who wrote this:
where you can see "General Product" in 1H2013 turns into won't ship till 2H and other logical jumps from a guy who open hates EMC and is paid to write for several of the smaller flash outfits.
Frankly, nobody innovates anymore. IBM bought Texas Ram SAN, HP bought 3PAR, Dell bought who cares, NetApp buys companies and takes 8 years if its software and kills it if it's hardware. The worse thing is that all of these startups are basically the same people, who don't want to become part of the big corporation so they take their buyout and start something new.....that basically is same as their previous startup. We had the folks in from companies bought by Dell with their new product, so new that they hadn't had a drive failure yet, but with their "new" "compelling" product where they stuck fusion io cards into a server and attached some jbod. It takes the bigger companies taking ideas from a few startups and combining them into a disruptive product.
Re: The market doesn't want unified storage? Really?
Yea, way to finally integrate software you guys purchased in 2004. How have IMC, Decru, Alacritus, and Topia's IP done?
And why would you ever come into this conversation with this angle? There's a reason NetApp bought Engenio, and it undermines any single point you just tried to make.
You need to learn from D from NetApp's playbook, cry "FUD" and unleash the dogs of marketingspeak.
I recently was visited by a storage company that had yet to have a hard drive failure (good thing too, no online spares and proactive sparing limited to SMART monitoring). Then there are the folks on my team dealing with Oracle ZFS, keeping them under 80% utilized while the system heat and softare raid kills an unusally large number of disks while getting the historically awful Oracle software support.
You get what you pay for.
Re: Not enough...
They're getting desperate to quote 3PAR for 14TB.
When large corporations that purchase hard drives by the trainful get a trainful of defective drives that company then has to spend resources and time replacing all of those drives that get installed.....and then they spend time and resources finding another supplier of enterprise quality drives.
When you don't have a meaningful SSD product and the industry has heavily bought off on using them.... then there's another meaningful supply you're not a part of.
Seagate spent a whole lot of money for maxtor and has nothing but manufacturing capacity for a crap product to show for it. Time for a lot of changes in management and vision.
They acquired a company acquisitions person FROM CA?????
Ok, first CA has a lot of successful Mainframe products that they bought from other companies. In addition, the name CA used to be synonymous with "pyramid scheme" as they bought software companies and sold them off peicemail (sp?).
So from a let's get someone who can buy companies, which has been sort of an issue for them of late, they found a someone from a company that can buy companies.
But this isn't a company that was good at integration. They were good at making a buck at the expense of butchering another company. How long did that spinnaker integration take? How's sales of Decru units? Is this the week the sales guy is trying to sell you Engenio so suddenly multiple product/code lines are bad?
NetApp needs help finding companies that compliment Ontap, embrace the waffles, and mainly help with VDI stuff, because outside of VDI NetApp's getting their butt handed to them these days.
Re: Nice graphs
Why include devices that don't use hardware raid or proactive sparing or has to run with an increased amount of spare space due to an unusually high drive failure rate? Or systems if you utilize more than 80% utilization switch to a different algorythm that severely affects performance.... well that would kill a couple others from that graph....
But it's really not that interesting, for all their hype, sales pushes, and marketing they're in the other category.
You never bought 3PAR because it was a barn burner, you bought 3PAR because it was THE innovator in things like wide striping, thin provisioning, and tons of other technologies that made it the industry innovator. That magic quadrant before HP bought 3PAR would have had HP closer to Oracle. It fit in a fantastic niche, better than average midrange with high end features.
Suggesting it's amazing because of it's SPC-1 scores is frankly terrible. While they didn't shortstroke nearly the amount Oracle or NetApp does (15 percent versus the others 50 percent or more), the read response time numbers as the load ramps up are terrible, the write's suffering less only because the entire box is configured as RAID-1.
But the fact is, once upon a time if you wanted to know what features the other vendors where putting into their arrays, you looked at 3PAR, but you don't anymore. And while everyone likes to point at HP and say, LOOK, they're screwing something else up, remember that they haven't innovated at all, which means the issues with moving forward predate HP.
Re: Both companies in trouble
Your single point of reference is fantastic, and completely opposite of my experience.
However, actual studies have been done, and Seagate has the highest failure rate of the industry. A lot of this is from their use of the liquid bearing drives from the purchase of Maxtor, but anyone who works in the enterprise space can tell you about Seagate firmware issues.
Re: For your HDS needs
I have to imagine with the VSP code running on it that they're going to price and position it as a VSP Lite, for those that want the virtualization features but don't want or need to purchase a full VSP (Think VMAXe or VNXe from EMC), or the folks that were going to integrate a HUS with a Bluearc.
The worry about midrange and virtualization is being able to push the data through the engine to the virtualized storage fast enough, so when you're doing it you're going to keep your local storage capacity down. The USP-V had limitations especially with cache that led to HDS telling people not to put databases on virtualized storage (not sure if it made it to the VSP)
For your HDS needs
Always stop by Nigel Poulton's blog:
The key points:
1. It's not AMS nor is it VSP for Hardware, it's based on a new custom ASIC with new controllers. Thus, HDS has 3 platforms now to manage and Nigel has serious concerns about scalability
2. "HDS are marketing this with a capacity sweet spot of between 20-180TB" which leads to why you would ever compare it to a VMAX. This box belongs against 3PAR, not a true high end enterprise (NIgel uses Midrange 1.5). Or, once again HDS is trying to invite comparison to the higher end product versus letting their sales guys sell where the product should be sold
3. Nigel points out that doing the NFS duties is a pair of BluArc NAS file servers, so this is "unified" just like an EMC VNX, or in other words, "unified marketing". Still going to be upgrading two different code sets.
Your missing the other big thing from what LANL and EMC have done with the new subsystem. Before, all computation had to be done and completed, then the nodes would then do the modeling (pretty video). Now, with the system in place, they can do computation and modeling at the same time, thus greatly reducing the time to run the "experiment"
The checkpoints now also occur at a much greater frequency than 4 hours, thus reducing how much time they lose when a node fail
I saw a demonstration where puppet discovered a number of servers from HP, Dell, and Cisco, and then with a couple of commands deployed full vmware clusters with vcenters in a couple of minutes with a cluster for each server type. Was amazing.
Was this paid for by NetApp? Seems like not 5 minutes after this annoucement every NetApp guy I know was talking about EMC pissing off it's partners.
Sans the fact that EMC has been increasing integration with partners with VSPEX and adapted a partner first sales strategy, what "partners" is this going to upset? Symantec? Hasn't had a good relationship with EMC in years and directly competes in the software side. Commvault? who?? And the fact of it is, with the fact that it doesn't integrate with actual Avamar, etc... that enterprises are still probably going to use their existing solution.
And as for price jacking notice VMWare just dropped the vram tax?
It's simply a way for the smaller companies to use functionality that should be there with a superior engine. Nobody complained that EMC was upsetting partners when ESX 4 high availability agent/process was a Legato product.
A lot of companies are moving to flash from java for their management interfaces. At a minimum it's just plain faster to render your management app and at a maximum you don't have to worry about when Oracle is going to show up demanding you pay for an enterprise license for a open source product they put out for free.
Re: Time Computers
Lexmark cartridges had a chip in them, for a while someone was manufacturing knockoffs and Lexmark sued them out of existance using the US DMCA to claim the company was bypassing the electronic protections on the chip.
Re: Nice try, big boy
In your link the flash card doing dedupe is going into the array "filer head". EMC's product goes into servers, thus the phrase "used in server flash caches"
NetApp has pretty much failed at any point to be an innovator, has been fairly obtuse about flash/ssd, and any attempt by them to purchase anything other than an also ran company will result in EMC deciding it needs a little more intellectual property.
"The company emphatically disagrees with flash as a storage tier"
They mean they can't get SSD to work in their arrays in any meaningful performance way. WHOOPS
Serious bonus points for leading with an API though, but every single all flash array out there is going to leverage it to be the flash tier that NetApp can't get to work.
I work at a company that's been heavily involved with Oracle and getting OVM working on Exalogic.
They've consistently missed milestones, our engineers have done most of the documentation while Oracle could not provide anything meaningful, and the product pales in comparison to any other Virtualization tool we use or evaluate when it comes to features and even the capability to move vms to other nodes.
The company bought some of the first exalogics off the line and we have zero production applications on the platform.
Re: completley missed the 2010 Olympics
Massive horrible coverage. NBC has had a stranglehold on it, just got a huge deal to have it pretty much forever, and universally everyone hates it.
Everything is tape delayed, nobody knows when anything will be on, hope you like swimming and gymnastics (women eat that shit up).
They have a total of 6 channels, the other day those six channels had:
2)Football (soccer for us)
3)That same soccer game
4)Basketball (game nobody cared about)
5)That same Basketball game
Don't even get us started on how horrible the announcers are.
Exadata is fast at queries...that involve lots of where statements, as long as you don't need to do a lot writes, as long as your front and back end aren't too busy with columnar compression CPU time, as long as your database isn't that big, as long as you have all the patches with data loss and unavability bugs loaded, as long as the software raid doesn't have two hard drive failures.
And then you get the HVAC bill. And then the Oracle maintenance bill.
Interestingly, right when Oracle hit the market full press with the second generation hardware, there was a bug in ASM with red hat where it would use the /dev instead of the multipath devices. Strange timing, performance limiting bug at that time.
As a Storage Admin Red Hat is one of my least favorite OS to deal with. Plus everyone I know who has dealt with XFS on Linux has tersely advised opensolaris instead. I just don't see using Red Hat as the front end for storage.
Re: Does QTM hold the patents... not DD
You're right, the original patent for the actual deduplication belongs to Quantum, Data Domain paid them for the license with stock during the original IPO of Data Domain. Data Domain said they held the patents for deduplication "in flight" versus post process.
Quantum was in trouble already when EMC bought DD and severed the agreement (EMC had given them a couple of cash infusions). Most likely a combination of having issues funding research and development and their sales capabilities.
NetApp: 70,481.804 GB of Unused Storage
3Par: 33,592.791 GB of Unused Storage
IBM: 38,979.496 GB of Unused Storage
Just keep that in mind. Great copy/paste of D from NetApp's blog though Chris. Although one thing about 50% markdown from list, this is a company with shrinking market share. Maybe part of that is pricing.
Nice throughput and grats on the 6 way cluster though. You guys should keep going with the IOPs in that testing, see where the actual performance fall off ends up.
NetApp had it's engineering working on a VTL product long before it tried to purchase Data Domain. I even got to see a demonstration of the product. It's either a)too bad they decided to cancel it or b)yet another example of why NetApp's software engineering group needs a revamp to get a boot up their ass to get things done (Hey, 8.1.1 finally is integrating technology from how many years ago).
When I was meeting with Data Domain long before EMC purchased them, one thing they constantly came back to was they were the only vendor that could do inflight deduplication because they had the patents. I wonder how many small companies trying to get into the market too got bullied out.
Oracle was also a high level sponsor while advertising the crap out of it's NAS and VM products all over Las Vegas, and the Oracle EMC relationship is very frosty. Just because a company pays to be part of the vendor show doesn't make it "Weird".
Of course EMC will sell you tape, they are a company that "sells" things and customers get what they want. Tape is not dead, never has been close, and "Tape Sucks" was the marketing of Data Domain prior to the purchase and EMC rode that for a bit to grab the market share the much smaller Data Domain couldn't get.
I'm not sure what your intent with the article was. Last minute dash to a deadline? Try to get your Stornext stock up? Very little to do with EMC actually, more of a simple comment on tape, but I guess trying to paint EMC as "wrong" gets you enough page hits not to have a failure of a opinion piece. If you get more desperate to get page hits maybe you can change the title to EMC no longer thinks tape sucks, but cups the balls since you've been on this weird 12 year old sex reference kick lately.
You guys are getting too hung up on acquisition price in your efforts to disqualify a beastly effort.
The point isn't that it's that you only get 4% of the capacity, it's that 4% of the capacity in a v7000 outperformed a frame they had to shortstroke the crap out of. If you honestly think another 96% of capacity is going to double the price of a storage frame these days then I have a large british landmark to sell you.
This is why SSD is such a disruptive product to storage, but it highlights the real thing you need to look at. It's still very expensive. So you have to have a system that can leverage your fast SSD and your slow SAS/SATA without great effect on your application. This is where you need to look long term. All flash is great, but you aren't going to buy an all flash v7000 if you can buy a hybrid v7000 that automatically moves data between tiers depending on performance/access needs.
In a few years, sure, all SSD. but there are a few years still.
(Also, since you all brought up price, acquisition price is one thing, operational expense is another. One of those systems is a lot hotter than the other. These things are more important than ever).
Sooo what was the previous market share of LSI/Engenio?
They still don't have cloud. They call what they have "cloud", but it fails to meet what clouds are on so many levels it hilarious.
Few years back, Oracle said cloud was a gimmick.
When nobody listened, and cloud became the driving force, Oracle apparently came out with.....a gimmick and called it "Cloud".
World's most expensive troll. One Rich Asshole Called Larry Ellison, indeed.
Nobody holds a grudge like EMC. The steep price cuts with VNX where as much to punish Dell as it was the start of the market share grab from NetApp.
Dell should have watched what happened with HP's storage portfolio divergence and learned that lesson.
I bought some NetApp stock last week. Their numbers aren't that bad, they just are having issues with this all happening during a bad time for the Market.
EMC has absolutely taken back some of NetApp's share. Management keeps trying to blame the sales market (I forget the terms used), but nobody is buying it. The next step with a continued fall is to change up the upper management, but in the end NetApp is a major playor and there's nobody in the industry that needs what NetApp has enough to buy them. Engenio was a superb purchase for them, not for the hole in NetApp's portfolio, but because it's a product OEM'd by enough companies to make it a solid steady income.
The real winner is all the EMC VNX and NetApp customers enjoying lower prices.
- Geek's Guide to Britain BT Tower is just a relic? Wrong: It relays 18,000hrs of telly daily
- Product Round-up Smartwatch face off: Pebble, MetaWatch and new hi-tech timepieces
- Review: Sony Xperia SP
- Geek's Guide to Britain The bunker at the end of the world - in Essex
- Dell's PC-on-a-stick landing in July: report