![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
Enterprise Grade Storage Recommendations
- Thread starter NobleX13
- Start date
CombatChrisNC
[H]ard|Gawd
- Joined
- Apr 3, 2013
- Messages
- 1,140
Are they setting any expectations for compression ratios?
None given to me yet, but I'm expecting to see the patch drop end of November/Beginning of December knowing how things normally go with them. Then wait a few weeks to see if any big issues occur because I'm always leery of new patches from vendors dropping with big feature additions
The one thing my inside buddy gave me for info is that there is a website being put out to upload logs to from your array which will tell you what sort of compression savings you would expect if you enabled it. Like the XIO data reduction tool, I'm betting everything is similar to XIO for the feature.
t_ski
Supreme [H]ardness
- Joined
- Jun 13, 2006
- Messages
- 7,506
I agree the GUI is much better. I hate the old VNX Unisphere stuff (especially with VNX1) relying on Java and poor security settings. I actually had some support tickets opened that turned out just to be stupid Java issues. The last guy got an earful from me about how EMC needs to get that crap figured out, because people can't go around using old Java and special browsers just to manage their stuff.
So far the Unity Unisphere has been baving very nicely, and even plays well with IE (default browser at work due to some web applications that require it).
Our VAR tried to change our config to get the price down, saying that the coming compression update would give us double the space, We told him we didn't want to rely on that due to it being 1. not out yet and 2. not guaranteed compression rates and made him requote. So when it comes out, we will turn it on and try it, and maybe we can move more servers on this than what we originally planned,
The only way your going to know what you dedupe and compression is, is by loading your data on it. Starting out, you should see at least 2:1. If you can.. try and separate the data as best as you can. Don't mix your oracle/sql and other stuff on the same pools as your vmware stuff. It'll throw your ratios off. EMCs wild claims of 10:1 or 25:1 are sales speak. It really depends on your data. They do have software that can do assessments to get a decent idea of what will do what. Your vmware stuff should compress fairly well. But you'd be surprised depending on what you have running in your vms.
t_ski
Supreme [H]ardness
- Joined
- Jun 13, 2006
- Messages
- 7,506
I will say that we also have a Data Domain 2500, and dedupe on that is very nice. I don't recall the figures off-hand, but I think it is somewhere around 24 or 25 to 1. I'd like to see that in the Unity, especially since we are doing VDI & Horizon View clones.
Still interested to see what benefit (if any) comes from the compression. I'm not expecting much.
Still interested to see what benefit (if any) comes from the compression. I'm not expecting much.
Child of Wonder
2[H]4U
- Joined
- May 22, 2006
- Messages
- 3,270
2:1 is probably a decent estimate for compression for a mixed workload environment. Results may vary obviously. I'm curious if it'll have any effect on performance on the Unity and VMAX platforms and if the IP came from the XtremIO line.
I was asking my Dell Emc partner about this same question today since I'm curious too. No"official" answer from him but the wink wink nudge nudge hinted at yes this is the case. Just have to wait for the patch and official details to confirm and get the details. I'm thinking the sales channel partners will be able to give us more info before the patch drops. Just depends on what NDA is there on the info
So we lost an SP last wednesday on our VNX5300. It keeps failing over to the other SP. Ticket was opened, and they have updated it and assigned it with Unisys. and so far nothing has happened. And it's been almost a week. Like I said, terrible support with EMC.
t_ski
Supreme [H]ardness
- Joined
- Jun 13, 2006
- Messages
- 7,506
What kind of service contract do you have? I always get same day or next day support, but that's the support we pay for. Do you have a vendor you work with, or are you working with an EMC salesman? I personally wouldn't have waited a week, but I would have called my sales team or SE and complained that there wasn't any action on the SR yet so that they can push the SR from the EMC side.
I would of been calling for an update and tracking the next day if I hadn't heard anything. Call and page the CE for an ETA
4 hour support 24x7. We still haven't gotten it. So I found out today, that Unisys basically screwed up again thinking it was an UPGRADE request so they ignored it because the tech is out of office. Now EMC Support is sending someone from Dallas to take care of it tomorrow. Reported 1 week ago, still not fixed. Basically found out the Control Station keeps rebooting itself.
We normally work with our EMC sales person, but because of all the issues we have, EMC gave us an EMC Customer Advocate, that usually costs the customer $50k a year, but we got her for free right now. She has been dealing with it the best she can.
t_ski
Supreme [H]ardness
- Joined
- Jun 13, 2006
- Messages
- 7,506
I would lean on her as much as you can, and make sure you voice your concerns during and after the SR. I usually get follow-up survey emails, and if I'm not happy about something they hear about it (usually when Remote Proactive starts and upgrade without calling me for permission, even though it was noted during the conversation for the SR).
If I had a similar issue and had not heard back in the four-hour window, I would have been calling for an update immediately and not waiting a week. I'm in healthcare in a 24/7 location. If something is down, business hours don't matter but patient care does.
Yep, I hear you. We are a medium sized bank, 6 billion sized. So, it can cause a major issue if we go down. What's even interesting...they were going to do the code upgrade on our VNX5300, and did the precheck, that is when they saw this issue. We got no alerts on it, it didn't send anything to ESRS either. LOL! So once they get this fixed they can do the code upgrade when we tell them. It was like pulling teeth on them, to force them to do it when we want them too.
I switched gigs recently.. my new shop has 4 450 all flash vmax arrays, bunch of xio, and DD. There isn't dedupe on the new vmax arrays just compression. Its very possible dedupe will come with an enginuity upgrade at some point. But compression has been there from around the 40k line up. The difference is compression is now off loaded to a special asic on the vmax instead of the compute engines. They started that process with first offloading at rest encryption to an asic to see how that worked out. It did well, and now they are expanding certain functionality. Unity is basically vnx3. The biggest unity arrays? They are using the same compute engines as the vmax arrays, so they have the same asics on board the controllers. With that said, I asked my reps and they said there is zero performance impact with compression on vmax today. I'm currently looking at a unity solution for tier 2 storage stuff. I can update you when I get more details. I'm getting wind of some new XIO hotness coming. I'm really hoping a code upgrade will enable the ability to inter mingle 20k, 40k and the rumored 80k brinks in the same cluster with the ability to migrate data between nodes so decomming older nodes is a snap. Rumor mill is blowing up, but using 6TB ssds in roughly the same form factor as the 40k bricks? That would be sweet.
You maaay come close, but to be safe, base all your planning off of 10:1 instead of 25:1. DD compression and dedupe process is post and not inline like xio. Being a post process, its not designed for speed hence why your clean can take quite while but your rates can be dedupe and compression rates can be pretty epic. If your array is doing nothing but 100% VDI, you'll see some stupid compression and dedupe numbers, but thats easy to do when most of your vms have 85 - 95% commonality. If you can with that array, try not to intermix data type. keep vdi in one pool, direct attach luns in a different pool, and db/oracle/sql stuff in a 3rd. You'll get better results. I know its not always possible to do that, but try if you can. If you can't, its not tragic, your rates will not be as good.
I've never had an experience like the one you've had, unless my support has lapsed on the array. If your support is valid, you need to call into support, transfer to clariion support, ask to speak with the mod, do the "I'm not getting off the phone till I get what I need, you will not put me on hold, and you will not transfer me". Yes it'll be a bunch of awkward mumbling, and sighs, but its worth it. You also need to call your sale team and have a meltdown on your sale team. Kick the ticket to your sales team, and make them run the issue. They can and will do that and will get you the results you need. It sounds like you are not melting down hard enough, because you can easily get that part if the parts depo is close by in 4 hours, or 12 if they have to fly it in. Don't do it via email, start making calls.
Child of Wonder
2[H]4U
- Joined
- May 22, 2006
- Messages
- 3,270
Having to scream at support and the sales team to get any resolution to a hardware issue? That's awful. How is this even within the realm of acceptable behavior for a hardware vendor?
Oh man, from bad to worse. Tech came out today, been there since 11am. I didn't get to have lunch, and we worked straight till 5pm. Sadly he ran into issues with the part we got in, and tried all he could do and his boss to get their support on the line to assist him. Well their support wouldn't even call him back. He waited 3 hours. So he now has to spend the night in a hotel, and will continue working on it tomorrow which I am out. My boss will have to handle it. He was so pissed off at his support it made me actually feel a little better seeing that they are going through the pains we go through all the time.
This is actually EMC that came out since Unisys didnt have anyone to get out to us.
Last edited:
emc clearly doesn't want your money. I wouldn't blame you one bit if you moved to a completely different solution. very unfortunate.
Naomi Goldberg
n00b
- Joined
- Mar 20, 2016
- Messages
- 1
You might find the user reviews on IT Central Station to be helpful if you want to avoid vendor FUD. NetApp FAS have particularly high user reviews when it comes to stability and technical support.
(disclosure: I work for IT Central Station)
(disclosure: I work for IT Central Station)
Honestly, with the larger enterprise SSD's, smaller footprint/power/cooling requirements, market drop in prices, and improvement in data services, it's rare where flash arrays aren't the more economical choice over spinning disk arrays. It's really not a performance sale anymore.
You have to have a really small environment, or some unusual workloads where a hybrid array is a fit nowadays.
As far as recommendations, I'll just say I work for one of the vendors mentioned in this thread, and my opinion would obviously be biased so I'll keep quiet. That being said, Gartner/IDC/etc studies are generally pretty accurate.
But if you happen to live in northern Florida and looking at making a data center decision, shoot me a PM
You have to have a really small environment, or some unusual workloads where a hybrid array is a fit nowadays.
As far as recommendations, I'll just say I work for one of the vendors mentioned in this thread, and my opinion would obviously be biased so I'll keep quiet. That being said, Gartner/IDC/etc studies are generally pretty accurate.
But if you happen to live in northern Florida and looking at making a data center decision, shoot me a PM
aaronspink
2[H]4U
- Joined
- Jun 7, 2004
- Messages
- 2,122
So true, about the only thing rust is useful for is T3/4 storage these days using 8-10TB drives purely due to cost savings. With things like P3520s reaching pricing parity with 15k drives, little point in having active storage in anything but flash.
One thing to consider..Until you get sued, and legal says you can't delete anything until the legal stuff is over, which can drag on years.. spinning disk is cheaper per gig then SSD. Whats cheaper then spinning disk? Tape. Both of which is needed for super long term cheap and deep storage. I had to learn this lesson the hard way.