Child of Wonder
2[H]4U
- Joined
- May 22, 2006
- Messages
- 3,270
![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
Just going to leave this here...
- Thread starter Child of Wonder
- Start date
peanuthead
Supreme [H]ardness
- Joined
- Feb 1, 2006
- Messages
- 4,699
Woohoo! This shall be interesting to say the least. I personally hope they develop the SMB market more.
DeChache
Supreme [H]ardness
- Joined
- Oct 30, 2005
- Messages
- 7,087
Child of Wonder
2[H]4U
- Joined
- May 22, 2006
- Messages
- 3,270
There is some crazy innovation going on with the up and coming storage vendors. Take advantage of the opportunity to shop around and learn more about what else is out there.
DeChache
Supreme [H]ardness
- Joined
- Oct 30, 2005
- Messages
- 7,087
I would love to. It's going to be a matter of convincing the rest of my team that can be to be nice stuck in their ways.
We have some many storage platforms its almost crazy.
VNX is a great product. VNXe? Meh.. I actually see Dell droping EqualLogic and Compellent in favor of VNX. If I wanted to really see crazy.. Drop VNX for XtremIO now that the 1.6+ TB enterprise drives are out with 2+ TB drives on the horrizon but keep pricing at the same level as VNX.
My biggest concern is Vmax road map. Vmax3 droppped less then a year ago, that platform will be around for atleast another 3 years, with 4+ years of extended support.
I see the 2 products being merged. In fact, Xtremio is now part of FAST-X.
Some marketing fluff..
I think the'll be merged because people want the performance of XtremIO, but they want the stability of Vmax. I was on a conf call with EMC that specifically said Vmax was build around spinning disks. Even if you fill a Vmax with nothing but SSDs, you'll still hover around an average of 1 - 3ms of response time. XtremIO being built around SSDs will hover at .5 - 1ms of response time. You could shrug at that kind of a difference, but in my environment (financial), its very important. But XtremIO does not have the same level of redundancy(today) as a V-Max.
I very recently finished doing a Vmax3 vs XtremIO vs Pure demo bake off. While XtremIO and Pure were neck and neck in performance, ease of use, setup, and administration, I settled on a Vmax3 due to internal redundancies, and maturity of the platform. Having the ability to virtulize a XtremIO array behind it in the future make it also very attractive.
Once EMC takes the best out of XtremIO and merges that special sauce into Vmax4, then we'll see a massive game changing shift in enterprise storage... Unless Dell ruins it.
Some marketing fluff..
I think the'll be merged because people want the performance of XtremIO, but they want the stability of Vmax. I was on a conf call with EMC that specifically said Vmax was build around spinning disks. Even if you fill a Vmax with nothing but SSDs, you'll still hover around an average of 1 - 3ms of response time. XtremIO being built around SSDs will hover at .5 - 1ms of response time. You could shrug at that kind of a difference, but in my environment (financial), its very important. But XtremIO does not have the same level of redundancy(today) as a V-Max.
I very recently finished doing a Vmax3 vs XtremIO vs Pure demo bake off. While XtremIO and Pure were neck and neck in performance, ease of use, setup, and administration, I settled on a Vmax3 due to internal redundancies, and maturity of the platform. Having the ability to virtulize a XtremIO array behind it in the future make it also very attractive.
Once EMC takes the best out of XtremIO and merges that special sauce into Vmax4, then we'll see a massive game changing shift in enterprise storage... Unless Dell ruins it.
DontPanic!
n00b
- Joined
- Sep 29, 2015
- Messages
- 10
In my personal opinion, usually smaller, more specialised enterprises deliver better products and services than "behemoth" ones (however they might have issues with global reach) - both EMC and Dell are already big ones, so I wonder if new "DMC" would be better...
Would be good to change the title of the post to more accurately reflect the content, CoW, you really should know better! 
The purchase price is pretty surprising to me. While EMC has struggled through recent quarters with massively falling spinning disk (VMAX/VNX) sales they are still posting a 24 billion dollar annual revenue. The purchase price is only 2.75 times revenue and that seems low.
The spinning disk lines must have much much worse forecasts than previously assumed.
The purchase price is pretty surprising to me. While EMC has struggled through recent quarters with massively falling spinning disk (VMAX/VNX) sales they are still posting a 24 billion dollar annual revenue. The purchase price is only 2.75 times revenue and that seems low.
The spinning disk lines must have much much worse forecasts than previously assumed.
Then what is the EMC all flash product going to be then?
That the big drawl right now is all flash arrays. If EMC doesn't get a flash array into the account, some other company certainly will try. Big drawl to XtremIO is the performance and the reduplication rate.They may offer solution you described as a spin off but to completely axe off XtremIO would be a mistake.
I think the biggest shake up will be to the higher end VNX model to make room for XtremIO and differentiate between mid market and enterprise product lines. Right now customer are comparing the two and making decision based on cost. Guess which one wins when cost is a factor?
Then what is the EMC all flash product going to be then?
They may offer solution you described as a spin off but to completely axe off XtremIO would be a mistake.
I think the biggest shake up will be to the higher end VNX model to make room for XtremIO and differentiate between mid market and enterprise product lines. Right now customer are comparing the two and making decision based on cost. Guess which one wins when cost is a factor?
There are close to 35 different all flash array vendors right now. 30 of them most likely will be bought out by all the major players. EMC already has XtremIO. Youre right, there is a big drawl towards XtremIO because of performance and deduplication rates. But XtremIO doesn't have the reliability of a VMax.
I'm not saying axe XtremIO.. I'm saying the'll take what makes XtremIO great and merge it into VMax down the road. XtremIO should also continue as a stand alone product.
High end VNX is the wild card. In my environment, I use higher end VNX's with about 500+ TBs of storage with a bunch of sata drives, some FC drives and a ton of flash cache. I use that for Dev/Test and QA environments, and I also FTS a lot of that storage behind a VMax. The cost is much cheaper to do that instead of buying a test/dev/qa vmax, but I don't have the performance demands as XtremIO. So the VNX is a great fit for me. Give it 5-7 years, and with the drive of SSDs crashing, meaning dollar per gig is getting much cheaper I see VNX becoming a very low end, high density archive solution.
Child of Wonder
2[H]4U
- Joined
- May 22, 2006
- Messages
- 3,270
There may be several AFAs out there today, but the trick is knowing the right questions to ask when comparing them because every AFA is not created equal and there is a ton of marketing FUD out there right now trying to confuse customers.
I work for an AFA vendor and when I talk to customers about AFAs, I try to arm them with the ability to ask the right questions so they can wade through the garbage being spewed at them from every direction. Vendor X comes in and talks smack about Vendor Y, Vendor Y claims they're technologically superior to Vendor Z, Vendor Z says they do everything Vendor X does but cheaper, etc.
- Which products are SSDs jammed into a platform designed for spinning disks or which are purpose built for flash?
- Was deduplication and compression bolted on? At what boundaries does it dedup and compress? Is it inline or post process? Does it hamper performance?
- Is all present and future software included?
- What does the upgrade path look like for capacity and performance?
- What is the lifecycle of the array? Do you have to forklift upgrade and purchase a wholly new array every few years?
- Is the product and company poised to take advantage of ever cheapening flash and new flash media? Will they get the same performance and reliability going from SLC to eMLC to cMLC to TLC as price pressures and capacity demands continue to increase?
- How much rack space, power, and cooling does each solution require?
- Does a component failure or upgrade decrease array performance and reliability forcing you to perform all maintenance during off hours or does the array perform equally well even as controllers or SSDs fail?
- Why was a particular architecture chosen (scale out vs scale up) and what limitations does it bring?
- And so on....
There's just so much more to it than "flash is fast and it dedups" and it's hard for customers to distinguish the real differentiators between each array.
The above questions are incredibily vaulable. When I was doing my Pure vs XtremIO POC, a lot of the above questions were asked. The answers were very surprising and eye opening. As an example Pure and XtremIO both do dedupe inline. However, if the Pure Array gets over saturated, it stops deduping until the load lightens and then it goes back and dedupes the data and any new data coming. Or, if you want to go from a 20K XtremIO brick, to a 40k brick, you have to rip out the 20s and replace it with 40s. Pure, you just swap out the engines. Which is better? Personal preference.
MrGuvernment
Fully [H]
- Joined
- Aug 3, 2004
- Messages
- 21,795
NetApp needs to buy out Nimble, Nimble really pushed the AFA market but because they were unheard of, people had trust issues going to them so they never took off like they should have.
Child of Wonder
2[H]4U
- Joined
- May 22, 2006
- Messages
- 3,270
Not exactly. Think of it as someone sitting at a desk and their job is to read large documents, file away the new ones, and discard ones already on file. If more documents begin coming across their desk faster than they can process, they'll begin to take a more cursory look at the incoming documents to determine if they should be filed or not rather than reading them line by line. Naturally this results in some documents being filed that didn't need to be. Once things slow back down, they'll revisit those filed documents to be sure they didn't miss anything.
I will admit, high bandwidth, large write IO isn't a Pure array's strong suit today but it's not a due to a hardware limitation but how the software was designed to ingest writes, which is something we continually improve. If a customer has a specific need around heavy writes or storing data that won't dedup/compress, I have no problem telling them our product isn't the best fit. I want every Pure customer to be happy and willing to tell others as much.
I could go into further detail or speak as to how a competitor does things, but I'll save that for a more one on one avenue.
Good information. Either you work for Pure or you've done your research. In my POC with pure and both XtremIO, I gotten virtually no compression and dedupe. This was with Oracle, but our data is very unique.
In regards to high bandwidth, large write IO? Most flash arrays in this department are about the same. AFAs really shine with small IO blocks, but very random IO and lots of it. I think as these products mature, large write IO will not be much of a problem. This goes back to my original thought.. there are 35 different AFA vendors out there.. They will gobble each other up, and then the main players(EMC, Netapp, and Hitachi) will buy what is left.
I was excited to see Pure go public last week. But they have a serious burn rate with their funds and haven't been a profitable company. I like the idea of yanking out the engines and replacing them with new ones after 3 years for free based on maintenance but that just can’t be sustainable. The “engines” are cheap.. They are rebranded Dell boxes. But most orgs will balk at a ridiculous maintenance bill. Most companies are trying bring down opex, not increase it. Sure Pure has a little special something something in the works and will no longer use Dell boxes, but that is a ways out. The fact that pure hasn’t been profitable, is a fairly new company with a new(non dell) tech stack and just went public? I’ll wait for everyone else to public beta that gear before I have another look at it. Honestly? I felt the same way about XtremIO.
In regards to high bandwidth, large write IO? Most flash arrays in this department are about the same. AFAs really shine with small IO blocks, but very random IO and lots of it. I think as these products mature, large write IO will not be much of a problem. This goes back to my original thought.. there are 35 different AFA vendors out there.. They will gobble each other up, and then the main players(EMC, Netapp, and Hitachi) will buy what is left.
I was excited to see Pure go public last week. But they have a serious burn rate with their funds and haven't been a profitable company. I like the idea of yanking out the engines and replacing them with new ones after 3 years for free based on maintenance but that just can’t be sustainable. The “engines” are cheap.. They are rebranded Dell boxes. But most orgs will balk at a ridiculous maintenance bill. Most companies are trying bring down opex, not increase it. Sure Pure has a little special something something in the works and will no longer use Dell boxes, but that is a ways out. The fact that pure hasn’t been profitable, is a fairly new company with a new(non dell) tech stack and just went public? I’ll wait for everyone else to public beta that gear before I have another look at it. Honestly? I felt the same way about XtremIO.
Child of Wonder
2[H]4U
- Joined
- May 22, 2006
- Messages
- 3,270
Yeah, some data just doesn't dedup or compress very well. Sometimes a hybrid array fits better in those instances unless a company doesn't mind paying for large amounts of usable flash.
Maybe. Maybe not. I think with how quickly technology is advancing and how game changing flash is, the smaller, more agile vendors have a leg up here. Already one of the "main players" themselves has been sold (EMC to Dell). Whether or not the little guys want to sell is up to them, but I think a really good product backed by a really good company can grow into a "main player."
The new m series Pure arrays released summer 2014 aren't re-branded Dell boxes. They're custom hardware purpose built to allow controller and SSD upgrades to occur within a chassis without needing to completely buy a whole new array every 3-5 years and moving all your data along with it. The controllers are just stateless compute and memory resources. You could attach the SSDs to a completely new pair of controllers and the data would all still be right there. That's what allows for easy in place upgrades to newer controllers. Takes about 90 minutes to make the swap to newer controllers with no loss in performance during the upgrade. How do they afford the "free" upgrades? Well, you have to pay for support, obviously, but think of how nice it is to not have to lug forward legacy gear when writing code. That represents a cost savings to Pure as well as agility to include new features.
The support costs are designed to decrease over time rather than increase (after all, flash media and computer gear tend to get cheaper over time) and the SSDs are warrantied for as long as you own them and have support.
Anyway, I'm getting way off topic here since I started this thread to talk about the Dell/EMC merger. Feel free to reach out privately if you want to talk more.
The good news is, you are saying exactly what my Pure guys are saying to me. Good to hear it and it validates what my local guys are saying as legit.
And yes, back on track. The big freak out I keep reading about is how they will handle the VMware aspect of it.
And yes, back on track. The big freak out I keep reading about is how they will handle the VMware aspect of it.
Let's be honest. Any of the true AFAs out there exceed probably 90% of customer workload I/O requirements. I try really hard to stay out of the "we flip bits this way, they flip bits that way" discussions. It's a waste of time these days. Go with the company, portfolio, and partner you trust the most. The rest is competitive BS.
Child of Wonder
2[H]4U
- Joined
- May 22, 2006
- Messages
- 3,270
Some of it really does matter beyond performance: support, scalability, etc. That being said, yes, most customers with their 1k-10k IOPS needs will easily fit into any AFA performance range and a great partner you can trust can steer a customer in the right direction.
kdh said:
My guess is Cisco is ejected from VCE, they purchase their own storage company, and release their own competing product. There's a lot of server sales to be had from vBlock and Dell isn't going to not butt in.
I hope that VMware can eventually spin away from Dell or, at the very least, be allowed to continue to be independent and function without direct Dell influence. A strong VMware pushes competition in the virtualization space and that's good for everyone.