Feeds

back to article SME storage challengers emerge one feature at a time

The storage market is in the middle of an evolution. To what, nobody knows, but a whole lot of people expect to get very rich between now and when we've got it all figured out. The only market that matters to most companies is the enterprise, because that's where the big margins are. One sales junket, lots of terabytes sold. …

COMMENTS

This topic is closed for new posts.
Anonymous Coward

Shouldn't there really be a disclaimer...

If you are writing adverts promoting yourself? I appreciate it takes a certain type of narcissism to write articles based on opinion rather than fact, but yours are beginning to get rather similar and rather boring and I’d rather be warned up front...

4
3
Bronze badge

Re: Shouldn't there really be a disclaimer...

Not sure I'd call that an advert. To me it reads that most IT guys will bring you a tried and tested solution based on common hardware while Big Trev will randomly pick from an assortment of new toys he's found and put together a unique design based on assumptions about how the technology will perform. He will then make it complex enough that you can't possibly replace him.

Perhaps I read it wrong though :)

I should add that for once I kind of agree with Mr Pott. It takes a completely different mentality to deal with smaller customers and I hate being asked to work in that kind of environment because it just feels like chaos after an enterprise customer.

2
0

This post has been deleted by its author

Gold badge

@AC re: adverts

I do believe you enjoy being incorrect. Why - exactly - would I write an advert for a service I have no interest in offering? My existing sysadmin clients are in maintenance mode. I have a negative interest in acquiring more and I am trying quite hard to extricate myself from most of the ones I do service.

Put bluntly: systems administration doesn't pay well enough. I'll probably be doing it for the rest of my life, but mostly as a favour to those specifically requesting my services and certainly not as a means of making a living. Systems administration hasn't paid the bills for nearly two years. Every dollar I make doing systems administration is 3 dollars I could have made doing any number of other activities.

Your personal prejudice only reveals your own limited worldview, AC. You know far less than you think, and I am increasingly unsure you think much at all. Cheers.

0
0
Gold badge

Re: Shouldn't there really be a disclaimer...

@Lusty actually you read it wrong. I don't think SME-focused MSPs will "pick from an assortment of new toys [they've] found and put together a unique design" at all. Instead, what I see is that they are standardizing on a new generation of providers that do everything SMEs require while offering direct margin support to the MSP crowd.

This is important, because there is a driver amongst MSPs not to give up margin to enterprise vendors as margins are getting thinner every day. (It's the reason, for example, I'm getting out of the game.) Because they are choosing a standardized set of equipment, they can keep spares to hand should something go wrong, but also refine in-house procedures to make things simple, easy and efficient.

I don't see anything complicated about deploying a Synology or Qnap-based solution into an SME. They're really easy to use. What they don't have, however, are "Synology certified supertechnician" programs, so finding a new consultant to replace the existing one would be hard.

SME-focused MSPs are not big on adding complexity. Perhaps even less so than enterprise-focused types. They handle dozens of customers where they can't offer too much attention to any single one. The last thing they need is complexity.

What's happening, however, is that there is an emerging dichotomy of providers. These SME-focused MSPs are choosing to provide hardware and software to their clients from different providers than the traditional enterprise vendors. I'm seeing it more and more with MSPs from around the world.

This makes it harder for an SME-focused MSP tech to break into the enterprise...but also for enterprise-focused techs to come down and support an SME. The gap here is bigger now than it ever has been, and it looks to be getting wider.

It's not a complexity issue, per se, though there is something to be said for the fact that SMEs tend to have a far larger total number of vendors supplying gear. (Because they can't afford to buy single-vendors stacks, and they tend to buy small/cheap cloud services from multiple providers.)

The issue is more one of a new generation of vendors being more than "good enough", vendors that enterprise techs would never give the time of day to. That makes for inter-professional jihads on the part of the differently focused sysadmins, but it also is creating a gap in product-focused skills and knowledge that we should all be aware of over the next few years.

This trend is something that has always been active at the lowest end, but I'm seeing it formalize and really pick up in the 50-500 seat range of SMEs. It is even expanding upwards in that I have seen things like Synology installed for high-IOPS production deployments in sites as large as 1000 seats. Eventually the squishy middle between enterprise and SME will start to get a lot more blurry and friction between the two camps will accelerate.

0
0
Flame

Re: Shouldn't there really be a disclaimer...

What pisses me off is when a consultant will suggest some new hardware/software that he/she has never touched. Then they want to learn the system while implementing it. I have no issue educating an employee who might stick around for a few years but I will not pay to educate a consultant. You are being hired because you say you know the system you submitted a bid for! Most small time consultants = setup wizard clicker.

If you are going to spend the money on a new system, pay a little extra and hire someone who knows what they are doing.

0
0
Bronze badge

You nailed it Trevor

For "small" SME's anyway. IT as a whole is just a PITA we would avoid if possible - it is often regarded as a cost without any benefit. Until it breaks.

0
0
Bronze badge

Re: You nailed it Trevor

I think in a very few years those customers will be on Azure and Office365 for the most part. This year MS will be binning the full Office install and licence so they will be forced to Office365 and the rest will probably follow in short order.

0
0
Gold badge

Re: You nailed it Trevor

Maybe, maybe not. The cloud doesn't actually offer a cost savings. That's the real problem. The price is going to have to come down quite a bit for there to be an advantage to SMEs. Remember that SMEs don't replace systems on a 2-year or 3-year cycle. It's closer to 6 years or 10 years between refreshes.

That's the whole reason Microsoft is pushing Office 365 and Azure so hard: they are addicted to subscription fees and they want to put a gun to the head of anyone and everyone to make sure that they pay up regularly. SMEs also aren't keen on change; Office 365 and Azure represent change you can't control and have zero say in. Windows 8 still weighs heavily on SME perception of Microsoft and that will haunt them for some time.

I have seem clients go into the cloud and then seen them come right back out in short order. I have seen clients go into the cloud and be happy with it. I have also seen clients avoid it like the plague. No one size fits all and no one solution is going to address all comers.

Can Microsoft convince SMEs to pay triple or more the cost of IT than they do now? That's the real question. When you start looking at the costs of IaaS and running a fleet of VMs on Azure the costs can be closer to 40x that of running your own infrastructure!

The cloud is not cheaper than buying and maintaining your own gear. That, right there, is the biggest issue with SME uptake. Microsoft is not likely to address that issue. They are far more likely to declare victory at some point where they feel they have enough customers and start turning the knobs on pricing in an Oracle-esque fashion. That won't end well.

0
0
Bronze badge

Re: You nailed it Trevor

I wasn't suggesting there were cost savings I was responding directly about small companies not wanting to manage the IT which Azure helps a lot with. It's certainly not cheaper and can be more expensive but I think it will start to take off one way or another.

0
0
Gold badge

Re: You nailed it Trevor

I know you didn't suggest there were cost savings. That's sort of where I was going with that. For SMEs, cost is king.

0
0
Bronze badge

HP's answer

to those little boxes is the HP Storevirtual VSA. Every Gen8 proliant comes with a free 1TB license (of course more capacity is available licensing based on capacity I think). Install it on your server hardware on top of VMware or Hyper-V and have full fault tolerance between servers with network RAID(real "HA"). Not only block replication but thin provisioning, online expansion, sub lun auto tiering, snapshots, live movement of data between storage systems, blah blah you know the rest. All the software features come enabled(I believe) out of the box. Don't need a full fledged SAN? Then don't buy one. I don't think the HP lefthand stuff is capable of online data migrations to upgrade to a 3PAR transparently(yet anyway) like you can with HP EVA..

Or if StoreVirtual is too complicated for some reason HP has StoreEasy as well.. again no SAN required. I'm certain Dell has equivalents to StoreEasy on their end, I believe StoreEasy uses Windows Storage server(and I think Dell uses that too on their entry level gear).

Not sure if IBM/Dell have something equivalent. I know NetApp has a VSA but last I checked it had no high availability(it did do replication I'm sure).

2
0
Gold badge

Re: HP's answer

@Nate Amsden Lefthand is for really small clients. Look at something like the Synology 1813+. I see these things everywhere. Little HA pairs of them littering the SME space. 2 units in RAIN 1 each with 2 SSDs (acceleration tier) and 6 Spinning disks in RAID 5 means 5x4TB usable at some pretty impressive IOPS for dirt cheap.

The RS3614xs+ is an example of the kind of gear I'm seeing put out by MSPs for far heavier workloads. I've never had a chance to play with one myself for an extended period, but I'm seeing them pop up.

The volume of data these things can shift and the IOPS they'll provide (hybrid storage arrays are a good thing) just make the Lefthand offer inadequate. That isn't to say Lefthand is bad for what it does, but it can't compare to VSAN or Maxta for the same job, it can't compare to Synology or Qnap for raw storage and it can't compare to any of them for price.

HP has good tech with Lefthand. They just have to figure out how to market it effectively to meet the growing challenges from proper storage SAN providers on the high end and upstart NAS/SAN providers on the low end.

1
2
Bronze badge

Re: HP's answer

I'm confused what you say by LeftHand is for really small clients, do you mean Synology or Qnap is for really small clients? It looks like StoreVirtual VSA scales to 50TB per instance which is quite a lot of storage for an entry level appliance. Performance I think varies depending on what kind of hardware your using. I have been told what might be called horror stories for performance on Lefthand with network RAID 5 (at least relative to 3PAR RAID 5 performance). Though having a shared nothing design does have a couple advantages over 3PAR(or any other SAN that is not shared nothing) from an simplicity/availability standpoint.

I mean as far as I know HP's own openstack cloud uses Lefthand storage (in part because it was the first HP storage platform to support Openstack, 3PAR support came about a year or two later).

I think StoreVirtual is a interesting product at least(I only used it once for a few minutes a couple years ago). I still believe HP should/will kill off the Lefthand hardware(which are just HP servers) in favor of entry level hardware arrays being based on 3PAR instead and make Lefthand exclusively a software only solution. They won't come out and say that though (much like they didn't come out and say they were killing off EVA for quite some time after they bought 3PAR). I think they will kill off the P2000 line as well for the same reason.

I agree their marketing needs some work I have talked directly with the marketing folks in HP storage on this topic on several occasions and they finally got my point last August I think, though I am not sure if they've done anything about it yet(I haven't seen marketing info for their stuff since in the event they updated anything), I wrote specifics on the conflicting messages HP is sending out about storage here last year:

http://www.techopsguys.com/2013/07/31/hp-storage-tech-day-other-store-products/

(jump to the Storevirtual section)

2
0
Gold badge

Re: HP's answer

First off, 50TB is nothing. It might be lots for really small clients, but it doesn't even begin to address most of my 50-seat clients, let alone those at 250 or 500 seats. Lefthand is for sub-100 seat setups, as far as I am concerned.

50TB maybe seems like a lot when you're using it only for VMDK storage. But then what are you using for CIFS? You could virtualize the file server and use Lefthand storage for that, but then you're back up against that 50TB wall. Or...get an appliance, right?

But why get Lefthand for VMDK and then a separate appliance for CIFS? Why not just get one widget that does both jobs, does it well and does it cheap?

Also: regarding performance...3par RAID 5 performance is inadequate as well. The compairaison isn't a bunch of rust spinning in a circle anymore. It's hybrid arrays. LSI's CacheCade or Microsoft's Tiered Storage Space on the DIY shelf, or Synology's SSD caching on the appliance shelf. Lefthand needs to be as fast as VSAN and Maxta (meaning taking rust and SSD to make superpowers) or it's just a toy.

A 32-node Maxta array can deliver amazing IOPS as well as insane capacity. You can also get a similar (though less dramatic on the IOPS side) effect from Synology storage. Those Synology arrays can top 600,000 IOPS now, and they cost a bent pittance.

Lefthand was cool in it's day. I really liked it. And in situations where all I need is an HA pair of servers to run a handful of smallish VMs without any real storage capacity requirements, Lefthand is fine.

For my customers who are needing 10-30K IOPS for their VMDKs and 150TB of 10K IOPS CIFS storage (about average for 50-seat setups I do,) it has so far proven to be far to expensive to attempt to make it work. There are just better alternatives at this time.

0
2

Re: HP's answer

Actually Trevor, your info is not completely right.. each instance/node of a Lefthand/StoreViretual VSA can handle 50TB , and you can create a 16-node cluster, so up to 800TB raw, or 400TB in Network RAID 10..

With adaptive optimization, all you need to do is have a few (preferrably 2 or more) SSD drives in each node and you easily get the high-IOPS you are looking for with Adaptive Optimisation (sub-LUN auto-tiering). And if you really want to have high performance (I'm talking +1.000.000 IOPS), you can use Fusion I/O cards instead of SSD drives.

And if you need even more performance & capacity you can create multiple clusters, and manage them all toghether as a single system with the Lefthand CMS....moving LUNs between clusters using Peer Motion if needed.. And to make it even more attractive, you get REAL HA with Network RAID 10 : no failover, no downtime for the storage (like Peer Persistence on 3Par).

Now on the 3par : Not really an SME box, but again same thing; using Adaptive Optimization you can balance IO's between SSD; SAS & Midline SAS on the fly.. and 3par has a bunch of other stuff : Thin Provisioning / reclamation, zero block removal, Peer Persistence (HA without storage downtime ! ) etc .... without influencing performance.

Both of them 3Par & Lefthand have hybrid capabilities ; mix and match SSD & SAS or Midline SAS drives.. 3par also has an SSD only array, the 7450 that is more than a match for everything out there..

And yes i am biased (i do work for the HP Storage division).

Cheers,

Frank

2
1
Gold badge

Re: HP's answer

I did not know about lefthand finally getting real capacity numbers. All documents I've read on the thing only mention 1TB, 4TB, 10TB and 50TB as potential storage sizes. Having real capacity makes it almost a real boy. That's good news!

I had heard of Adaptive Optimisation, but had not realized it was at block-level. I understood it to be at LUN level, and thus utterly useless. Of course, it only kicks in at the 10 TB licensing level which doesn't address the price sensitivity of SMEs in the slightest.

From what I have been told - admittedly it's been a good long while since I've put 3Par in - the hybrid capabilities of the modern arrays are not that great. Certainly not Tintri standard, but many have said to me that they've seen faster acceleration off the things is really not that much better than Synology can deliver, despite costing three virgins and a tropical island more.

All of which brings me to the issue you very carefully failed to address: price. As mentioned in the piece, one of the biggest reasons that MSPs put these solutions into the shops of their clients is that the MSPs themselves get to keep the margin, not HP, Dell or so forth.

HP may have the ability to match or exceed Synology. I'd bloody well hope so, considering the size of the organizations and the fact that HP supplies real SANs to enterprises! But what's under the microscope here is the lowest-end, entry-level offerings of HP versus Synology (or Qnap or Netgear or any of the other up-and-coming entrants.)

"You can do everything these guys can on our entry level equipment" is a bunch of hooey because it very carefully omits the part where enabling all this stuff is additional licensing. Thus you are climbing out of entry level and getting into mid range, the prices decoupling from the raw hardware and sailing off out into orbit. This is why the MSPs I have been talking to say "yeah, we love HP, they make the best servers on the planet, but we can't sell them anymore."

That's before we get into the fact that HP doesn't let third parties who aren't part of the handshake club and paying ransom money to the mafia membership to the partner program service units. This HP make sure to vacuum up that margin from MSPs as well.

Then we can talk about using third party drives in systems versus the cost of HP-branded drives. Let's have that talk, too, shall we?

Look: HP makes great stuff. Phenomenal stuff, actually. A cut above virtually many and on par with most of the others. I don't think anyone will dispute that.

The question here is about value for money, competitiveness when compared to up-and-comers (like Maxta, Nutanix, SimpliVity, VSAN and so on and so forth) as well as the key, critical one: value for money.

For MSPs deploying to their customers, there are additional considerations. Do the MSPs get margin support? Do they retain the customer relationship? Are they forced to move upmarket into an increasingly crowded sub-enterprise space in order to be able to find clients who can afford to buy the gear?

Now, that being said...if HP has finally changed their ways, started commoditsing featuresets, begun to play nice with SME-focused MSPs and otherwise made a real effort to be competitive with the new crowd, that's fantastic. I'd love to do a followup piece and showcase how HP is responding to these folks and intends to make strong gains in the SME space. I'd sure as hell love to be putting HP stuff in place of ust about anyone else at my customer sites, and I could point you at somewhere in the neighborhood of two hundred other SME-focused MSPs that feel the same.

I'd also love to give the newest Lefthand a real go. If HP is convinced it has evolved to the point of no longer being a toy, it'd be worth trying out. Besides, the virtual SAN space needs some healthy competition. Margins need to go down so that we can really start commoditising hardware SAN vendors. I'd be quite happy if HP were actually taking part in that.

1
2
Anonymous Coward

Shurely shome mishtake

"Also: regarding performance...3par RAID 5 performance is inadequate as well."

Sorry but the baby 3PAR would leave one of these boxes in the dust just with spinning disk, with all SSD or a hybrid configuration it would crush it from a IOps performance perspective.

Take a look at the Synology specs, the 600,000 IOps is pure fantasy. They're sequential, 100% read and no mention of working set size, meaning we did it all from cache. For some more realistic numbers look at the the random IOps which I assume are also best case. Those numbers are not even close to the 600K figure and aren't even particularly impressive.

Don't get me wrong this looks a decent box for the money, but you can't in all seriousness compare this to widely deployed field proven enterprise kit.

1
1

This post has been deleted by its author

Anonymous Coward

50TB's is nothing, yet 10TB's is to much for price sensitive SME's ? You can't have it both ways.

"You can do everything these guys can on our entry level equipment" is a bunch of hooey because it very carefully omits the part where enabling all this stuff is additional licensing.

The lefthand / Storevirtual kit and VSA licencing is also all incllusive and you can run the VSA (which is also the most mature and feature rich available) on any vendors server and storage hardware. It also provides full interoperability with the hardware StoreVirtual and is priced very competitively against similar VSA products.

1
1
Anonymous Coward

Research

You''re suggesting other products are toys or inadequate in some way based on rumour, lack of research and out of date information. Given you''re in the relatively unique position here of publishing articles, which like it or not carries a level influence you should at least do your homework first.

1
1
Gold badge

"50TB's is nothing, yet 10TB's is to much for price sensitive SME's ? You can't have it both ways."

50TB is nothing; the thing has to be able to scale way past that cheaply. Simultaneously, what is charged for the 10TB tier from HP is too much for entry-level stuff. A lot of SMEs need to be able to start off at something small, like 2TB usable but expand rapidly with capital costs that don't exceed the cost of the disks themsevles by all that much.

I am not saying Synology are perfect here either. Synology's approach for their gear still doesn't allow you to pack many disks into the initial unit before you have to go buy an expansion device. I'd far rather see the primary unit be a 24-drive device with the ability to lash on two more 24-disk JBODs.

The biggest reason for this is that SMEs are terrible at organizing their data. They just hang on to everything forever. They're price sensitive, so they sure as hell aren't going to archive any of that to the cloud, nor are they likely to buy tiering or archiving software. They will buy physical hardware because that's tangible. They can touch what they're paying for.

Synology, Qnap and others are on the right path to make fast, bulk storage truly cheap and commoditised...but even they aren't there yet.

1
2
Gold badge

Re: Research

"You''re suggesting other products are toys or inadequate in some way based on rumour, lack of research and out of date information."

They are inadequate, for the target market I described. HP's 3Par arrays, with all the knobs turned to 11, butcher anything that Synology, Qnap or anyone else could field. But doing so moves the cost of the device right out into a completely different category. You aren't comparing like for like at that point.

You are quite quick to be dismissive of the Synology units from a performance perspective (and to lay into me) but you haven't done much of any research here. You're just latching onto anything that will make that the brand you've taken to heart look good. I have no such attachments. To Synology or anyone else. (Except Ninite.)

Let's look at those Synology boxes, shall we? So in order to achieve "stupid IOPS" Synology is using a system filled up with Intel 520 Series 240GB 2.5" SSDs. That's right, ladies and gentlemen, a bucket of old, outdated, slow and dirt cheap SSDs. Ones I am sure were really super-duper-cool when the original test was run, but now don't even come close to what you can get, even out of consumer gear. So what do we get with this box 'o rocks?

Sequential Read: 622,000 IOPS.

Sequential Write: 330,000 IOPS.

Random Read: 333,000 IOPS.

Random Write: 163,000 IOPS.

Now, I know from experience that Samsung's 480 Pro consumer SATA SSD will crush the Intel 520 240GB SSDs like a bug, write performance wise. Even newer consumer SSDs don't get much better read speed, but they do get far better write speed. Based on a little maths I expect that if you reran the test on that hardware with modern consumer SSDs your sums would look a heck of a lot more like the following:

Sequential Read: 600,000 IOPS.

Sequential Write: 500,000 IOPS.

4K Random Read: 500,000 IOPS.

4K Random Write: 300,000 IOPS.

80/20 Read Write: 550,000 IOPS.

Now, I haven't done the test yet on this myself, but this is based on some educated guessing and a lot of experience. I am currently trying to scrape together enough equipment to prove the numbers above, but I've very confident that they're roughly accurate for modern consumer SSDs.

Your SSDs matter. You won't get that kind of random (or sequential) write from, say, the Micron 960GB units. The controller is a bucket of wet sick for write speeds and does this bizarre thing where it spikes high then runs for a while at half speed. Using a Samsung 480EVO 1TB, however, you'll get a consistently high write speed. It won't be near as high as the read speeds, naturally, but way better than the Intel 240!

An HP unit filled with enterprise-class SSDs would probably be able to take the Synology system. It would cost half a continent's worth of virgins - whilst delivering bupkis in terms of capacity - but it might just get there.

Speed for the price, however? fugeddaboutit. Capacity for the price? Double fuggedabout it. The Synology (and competitor) units are a way cheaper path to falling-all-over-yourself-stupid IOPS and great capacity. Using off-the-shelf consumer drives that you can buy anywhere. It's a hell of a lot more flexible and a hell of a lot cheaper to achieve just about any goal you want...below a certain threshold.

There's a point where HP will absolutely wreck Synology. For that matter, I'd argue HP could probably wreck Dell, give NetApp some bad dreams and has EMC checking over their shoulder on a regular basis. That's way out there, big time in the far off la-la land of "more than the total gross turnover of the company spent of IT infrastructure."

As for Lefthand...I honestly and truly believe it's been surpassed by it's competition. I'd put Maxta up against it any day, to say nothing of Nutanix or SimpliVity. Not that I personally consider the latter two to be "SME"-compatible in any way, but they'll get you where you need to go cheaply and quickly if you're talking about a small CSP.

Look: I like HP. I even - despite what you may think - like Lefthand. I have a soft spot for the thing that I really shouldn't have, since it was so much better than VMware's VSA for so long and a reasonable bit cheaper. As a sysadmin, I love HP's servers and storage because the design of HPs gear is just so much better than anyone else.

HP servers an honest-to-$deity pleasure to work with. Cables are routed nicely, the tools you need are magnetically attached to the back...every detail on an HP server has been thought through for ultimate serviceability. I wish every server could be an HP server.

Synology units, by contrast, are nasty things with sharp edges and densely packed insides that were never designed to be serviced. If I were buying second hand, out-of-warantee stuff for my lab then I would never buy Synology, Qnap or any of these guys. I'd buy HP.

But let's face facts, here, shall we? HP can't match the capacity or performance per dollar of these up-and-coming companies. Synology, Qnap, etc are getting better and better all the time and they are eating the low end.

At the higher end you have this rapidly evolving storage field where there are literally dozens of companies form Maxta and Atlantis to Proximal and Pernix to Nutanix and SimpliVity all working their asses off to commoditise "proper" enterprise storage. They are dragging margins down and offering...(drum roll)...better capacity per dollar and/or better performance per dollar. All without compromising reliability.

HP isn't adapting fast enough. For that matter, neither are most of the big storage heavies. That's normal. It's also normal for those who work for Cisco (or owe their livelihoods to it) to get really angry when someone points out that Arista is kicking ass. Same thing applies here.

I've done my homework. I continue to do my homework every day. Sometimes I'm wrong, sometimes I've missed the existence of a company or product. Sometimes I'm even right.

Here, now, Lefthand isn't priced well enough to match Synology on capacity. Depending on how you build the things it might match on performance, maybe, assuming that Lefthand has improved enough in the last update set to match Maxta for most things.

In which case I'd just buy Maxta for my own servers because it's cheaper. That's assuming I even want converged infrastructure. I might just want a great big ball of CIFS, in which case any converged infrastructure play on the market today is bad value for money compared to Synology, Qnap, etc.

3Par can't go up against Synology on speed for dollar, to the point where matching the Synology would be prohibitively expensive. 3Par is better designed for serviceability. Which you can't have done unless you pay HP's ransom. 3Par is more feature rich.

None of this - none of it - addresses the issue that was actually discussed in the article itself: that of MSPs choosing up-and-coming storage vendors because those vendors A) provide margin support and B) allow the MSP to control the customer relationship.

It's about money. All of it. Price per capacity, price per IOPS, margin available to the MSP, servicing margin available to the MSP...it's all about the money. HP isn't prepared for this new, commoditised world. They aren't ready for the brave new universe where storage margins look like x86 server margins.

Soon, IBM will sell of their storage division. Probably to Lenovo. HP will be sitting there - again - with an offering way above the new rates the market has settled upon. HP needs to shit or get off the pot. Decide if it is going to be the "Apple of enterprise tech", charging $virgins and delivering quality, or if they are going to play in the muck with the mass-market people.

So far, I see no evidence of the latter, and I see a whole bunch of companies trying out the former. HP has a lot of competition if it wants to be "the Apple of enterprise tech". And the "mass-market muck" are getting more than "good enough" while being cheap enough to entice.

I'm not saying HP is crap. I'm saying HP hasn't committed to the evolving SME space and it shows. They are still in the process of deciding who they want to be. Meanwhile, the vultures are circling.

Storage just ain't the easy money it used to be.

1
2
Anonymous Coward

Re: Research

I wasn't dismissive of the Synology units at all, on the contrary I said they looked decent boxes for the money. What I tried to point out was the performance numbers you were touting are marketing numbers.

The product specs suggest insane 600K + IOPs for sequential reads only, the more realistic figure of random IOps is way way lower at 33K IOps. You added an extra 0 to your numbers by mistake, but looking at the way they lay out the numbers with the decimal points and no commas I can see why that was easy to do.....As someone who works with storage on a regular basis I could tell at a glance your numbers didn't add up for random especially when you begin to add writes into the mix. Which unfortunately means your projected numbers were also way off which could be a very costly mistake, especially if you're providing performance guarantees.

http://www.synology.com/en-global/products/performance/#xs_plus_series

Sequential IOPS (10GbE VLAN) Reads 622,848 IOps

Sequential IOPS (10GbE VLAN) Writes 330,689 IOps

Consists of transferring 16GB file continuously to/from 8 PCs for 3 minutes.block size 32KB (cache friendly working set)

Random IOPS (10GbE VLAN) Reads 33,389 IOps

Random IOPS (10GbE VLAN) Writes 16,382 IOps

4KB random IOps (cache hostile)

The other thing I pointed out as did others was that you were off in your perception of other products capabilities and despite your extremely long response that appears to remain the case.

I don't think there's much point arguing about the price if your sole scoring mechanism is bang for buck then you can buy the equivalent from anyone and assume product parity. But storage is more than just performance and that's where the pricing kicks in, because H/A, data integrity, platform stability, qualification and support etc, in fact the whole RAS package costs in R&D dollars. It isn't just about vendors gouging customers it's about engineered solutions rather than just lumping some components together, which TBH any number of assemblers can put together these days. If that fits your clients business model then so be it, but you can't directly compare the two solutions.

BTW the StoreVirtual VSA node license comes in 1TB (Ships FOC with HP servers), 4TB, 10TB and 50TB versions and you can upgrade via a license key between capacity points as you grow. Since you can roll your own on anyone's hardware, the price / performance mix is up to you.

1
1
Gold badge

Re: Research

To start off with, it doesn't look like you actually read my responses, because you seem to be very selective about what you actually respond to, or even acknowledge is said. Things like working out the actual numbers, discussing the differences in the SSDs used and acknowledgment that "marketing numbers" age as the tech underneath moves on.

I also like the part where you totally gloss over not only the evolution of SSDs, but enterprise versus consumer SSD usage in Synology and HP arrays. Or the ability to use drives from any old vendor rather than HP-mandated stuff. Irrelevant, I take it. Easier to attack the dated marketing figured of a company with eleventeen squillion times less budget to get tests run.

It doesn't strike you as ridiculous to be arguing over how much faster than the speed of sound the jet can go when what the commuter wants is to get from Edmonton to Vancouver faster than driving and without dying? Compare like for like drives in both systems. Maybe HP goes faster, but I personally doubt it. Certainly not enough to justify the price delta!

"I don't think there's much point arguing about the price if your sole scoring mechanism is bang for buck then you can buy the equivalent from anyone and assume product parity."

When you are a price-constrained SME bang for buck is the sole scoring point. When you're an MSP trying desperately to eek out margin, bang for buck is the sole scoring point. Or did you not actually read the article and the very specific target markets it was addressing?

"But storage is more than just performance and that's where the pricing kicks in"

For some segments, storage is more than performance and capacity. Don't paint all niches with one brush.

"because H/A, data integrity"

Both of these are not commodities. As pointed out, available even from the likes of QNAP or Synology. It's not a marketing blurb anymore, it's a tickbox feature that must exist as a bare minimum or you don't even get to play the game. Next.

"platform stability"

Vague FUD. Commodity gear such as Synology is perfectly stable. Tintri, Windows, Nexenta, Maxta, Nutanix, SumpliVity, ILIO, and on and on and on and on. They're all stable. Stability is the norm, not a rarity.

"qualification"

Qualification in the scenario discussed in the article is generally done by the MSP. That way they know what they're deploying and they get to milk that minor amount of effort for several years' margin, instead of the vendor. Which was kind of the point of the article.

"and support etc,"

Support? Support?!? Don't you mean ransom? Because you can't support HP equipment unless you stump up a ransom to be a qualified vendor. Oh, you mean the customer gets support? Well why is it better from HP than an MSP? HP offers 4-hour enterprise support? MSP offers "my cell phone is a taser attached to my testicles that rolls me out of bed and a drive across the city is 1 hr tops" support. Spare parts? Synologys are so cheap the MSP can afford to keep entire spare units on the the shelf and still make a profit, because they aren't paying the "support" to the vendor.

Is that optimal in all cases? Hell no. Is it increasingly common? Yes. There isn't a lot of breathing room for MSPs and this is what they are doing. So I talked about that. It's also part and parcel of where the discussion about "who owns the customer relationship" comes in, which you have been so very carefully avoiding.

"in fact the whole RAS package costs in R&D dollars."

Yes, I agree. The point here, however, is that the MSP would rather spend the R&D dollars and the reap the rewards. They take the margin, offer storage at a lower cost, both the MSP and the customer win.

"It isn't just about vendors gouging customers it's about engineered solutions rather than just lumping some components together, which TBH any number of assemblers can put together these days."

Arrogance much? So people "lumping some components together" and then taking the time to test them, qualify them and build a support infrastructure around that product just aren't as good as you? Even if their products stand the test of time, achieve the design goals, meet with customer expectations and reach a price point that works for all involved? Wow.

"If that fits your clients business model then so be it,"

Yo, back to the article, where I was talking about SME-focused SMEs and the SMEs they serve. Yoo-hoo!

"but you can't directly compare the two solutions."

Why not? What matters is fitness for purpose. Absolutely nothing matters beyond that. That includes everything from features to ecosystem to - you guessed it - price. Which means I can - and will - compare anything to anything else for a given niche to determine the fitness for purpose.

If you need to fly from Edmonton to Vancouver and your criteria are "get there safely", "get there within 3 hours" and "be affordable" you don't need to use a 737 or a Dreamliner to do the job. A Dash-8 will do.

A 737 or a Dreamliner would be more comfortable...but given that comfort wasn't a decision criteria then it's fitness for purpose is exactly the same as the Dash-8. Comfort being a luxury consideration would only be a deciding factor if the cost difference between the Dash-8 and the 737 or the Dreamliner were negligible or non-existant. All planes are fit for purpose.

So I can compare the Dash-8 to the 737 to the Dreamliner for this requirement set, despite them being completely different classes of plane.

Thanks for participating in the forums, and I hope you have a great day. Cheers!

1
2
Anonymous Coward

Is that the best you can do ?

I see you completely ignored the fact you dropped an absolutely massive bollock on your calculated I/O numbers :-) those aren't Synologies numbers they're your misinterpretation of them. Forget SSD advances because they aren't relevant to your numbers, in fact you don't know enough about storage to spot not only that those were cached sequential numbers, but also that your random numbers were impossible given those numbers. Whilst we're at it you were the one spouting off about IOps numbers vs the bigger vendors so I don't see how your speed of sound or jet analogy is relevant to my comments.

But never mind, just trash talk about something else, throw a few insults out and maybe no one will notice you were out on the IOps numbers by a few hundred thousand. As you said they were all "based on some educated guessing and a lot of experience" or maybe not so much of either given you made such a huuuuuge schoolboy error, despite the hand waving it hasn't really done much for your credibility. Good luck with any storage consultancy because your customers will need it, but do keep up the infomercials for your mates, it's comical reading your pseudo science.

Cheers

1
2
Gold badge

Of course you don't see how any criticism is relevant to you. I wouldn't expect otherwise. My numbers were calculated mostly by bunging a smaller number of EVO SSDs into a smaller Synology and extrapolating upwards. Maybe they're off, maybe they're not. They don't seem all that out to me, based on my own testing.

A Samsung EVO 1TB will give you around 22K write IOPS at 4K, after eight hours of preconditioning. You can put 12 of things into the unit. 12x22 = 264L IOPS. The unit can handle 2 expansions, neither of which will be able to deliver enough throughput to really exercise the drives they contain. (The expansion ports are just Infiniband.) Put the local 12 SSDs + the max realistic IOPS you'd deliver through the two expansion bays and 300K random write IOPS at 4K doesn't look unrealistic at all.

In terms of raw numbers, it's rather low. All 36 drives at 22K IOPS would theoretically deliver 792K IOPS. Not going to happen. Putting the throughput of the expansion ports to one side for a moment, a single Xeon E3 is a little anemic for that kind of work. Not to mention that doing this in software is going to introduce some pretty massive overhead. Then you have to crank it out over ethernet...

When I test a Samsung EVO 1TB on it's own I get 22K. 4 of the things should give me 88K. Putting four of them in a modern Synology instead gives me ~40K. So, applying the same (very rough) maths to the 792K peak theoretical figure for those drives (divide by two and round down a little) I come to 300K write IOPS at 4K.

It's inexact - because I don't happen to have 36 EVO SSDs, let alone the Synology units and expanders in question - but it's more than enough to meet the needs of an SME. That's a pretty sexy Dash-8.

I am mostly guessing that the reason the per-drive performance is functionally cut in half is because when I am testing a single drive I can maintain a queue depth of at least 4. With a Samsung EVO the write IOPS seem to drop off dramatically below a queue depth of 4. The read IOPS have a sharper curve, seemingly optimal around 32. That's conjecture, and at some point when I have free time I actually do intent to test that hypothesis more thoroughly with the Synology units I have to hand.

What really matters here is if Synology was using the expansion bays during testing. If they weren't then they were getting 13.5K IOPS per SSD out of the Intel 520 240GB units. That's in line with what I would see from one of those SSDs post-burn in.

If you apply my "add 'em up, divide by two" to this you could extrapolate that the raw 4K IOPS of the disks was 27K. That's far higher than I have seen from a post-burn-in Intel 520 240GB, but about right for fresh-out-of-the-box.

If Synology was using expansion units for that test then things change. With 36 SSDs that's about 4.5K IOPS per drive which is LOLWUT-class bad. I can't believe for a second that's the case, simply because their smaller units can cheerfully extract 10-12K write IOPS at 4K from the EVOs. Why would the same OS on more powerful hardware suddenly perform half as well?

So based on the above I would expect that if I built a fully complete Synology RS3614xs+ with two RX1214/RX1214RP expansions filled with 36 Samsung EVO 1TB SSDs there is no reason I should not expect to get 300,000 write IOPS at 4K, especially when the theoretical deliverable IOPS (based on a measured 22K per disk) is 792K.

One thing I will never get about storage people - and this is something I am noticing seems to be pervasive throughout the industry, from the smallest startup to the largest entrenched player - is why every storage nerd and their dog thinks that their pet storage outfit is magically one size that fits all. If there are 10 devices that are fit for purpose it makes rational sense to choose the cheapest of the lot...but "fit for purpose" seems to have gotten lost amongst a rising knell of religious fervor.

I will never understand this. It's just a piece of technology. This one, that one, the next one...who cares? It's relevant today and old news tomorrow. I will never understand what makes someone invest their emotions into a corporate brand or - even more bizarrely - a specific product from a specific vendor.

No matter how you might wish to insult me, a bunch of tripe int he comments isn't going to change that fact that I just don't care who wins or loses these little battles. HP, Synology, Dell, Qnap...they're all completely and utterly irrelevant to me. Interchangeable, replaceable, faceless, soulless corporations. Each and every one of them as irrelevant to me as the next.

None of these companies is ever going to make me rich. None of them will make my dreams come true. I can't imagine what bit of computer equipment I could possibly want that I don't already have. I have enough kitch to start my own trade show. I loathe traveling so junkets aren't exactly something that makes me happy.

No. I'm a miserable, jaded, cynical old git...and I like it that way. HP, Synology and all the rest could go bankrupt tomorrow, I wouldn't give a bent damn. There will always be someone else to take their place, and life will move on.

Value for dollar is the key...and it always intrigues me who is vehemently opposed to the concept, and which companies they champion.

Cheers.

0
2
Anonymous Coward

Syno gear

Has to be said, I've got a DS214+ here purring away under my desk, and it's genuinely impressed me with what it can do. For a boggo sub-30seat SME, I don't see why one of these with a couple of USB backup disks couldn't handle most of their data handling (as most small companies really just need a Samba share with user rights assigned and little more).

It'll also do print sharing and other little things. But it'll also do almost all of the softwarey things (replication, iSCSI, NIC teaming etc) that the monster rackmount models will, and this thing costs under £300 empty.

Want a VM server? Got an underutilised Windows 2008 server with VT-X instructions on it? Throw a couple of disks in pretty much any Synology NAS and bosh, instant iSCSI so you can run a test virtualisation environment without fecking about with the server itself, bringing it down to add seperate physical disks so that the VMs won't hit performance of SQL instances etc.

The MSP I work for do a fair few basic Windows Server installs which could very, very easily be replaced with midrange Syno units at well under 75% of the cost of a proper, simple Windows Server box (my one, including disks, was under £500) with comparable performance, and I reckon that could get us more clients, including maintenance clients who pay a subscription to us for rapid response and all that.

I mean, if you can get a reliable filer with good expandabilty and scaleout (And this goes for QNAP stuff too, from what I understand) then why not? If you need an Exchange server (or other vendor-specific OS based service) further down the line, you can get one. Until then? One of these devices will do just fine...

1
0
This topic is closed for new posts.