2014-02-06 06:03 AM
I'm sure their are success stories and horror stories both ways here, but i'm looking for some real world experiences/gotchas etc.
I understand Nimble is iSCSI only for now and that does play a part, but in our world this isn't a game breaker.
Anyone out their make the conversion to Nimble and are now sorry they did? Anyone have any concerns about Nimble as a company and their support organization? The thing i really dislike about Netapp is you have to buy the software each time you change controllers.... This really stinks....
2014-02-07 10:25 AM
this is from a client that just put in a 3220 with SATA and flash cache and another with all SAS..
"The migration is going well, after this weekend I should be 95% done. I can already tell from the tests I have run this will be a much more resilient system than our previous endeavour. I was also very surprised to see the 3220 with SATA and flash cache blows the doors off every other array I've tested, including the Nimble and the 3220 with 20 10K 900 spindles. I was also quite shocked to see how terrible our Equallogic in Toronto performs. NetApp SIO is a great tool for testing this stuff out and I wish I had known about it sooner."
2014-02-07 10:38 AM
Hi George -
I work in Competitive Intelligence here at NetApp, and you are correct, we can sit all day and swap horror stories. I can assure you that it won't help you rationalize your decision you've made. As you can see from the silence, we are a bit more polite at NetApp. It is the culture.
There are a few considerations an enterprise of any scope needs to make when purchasing a foundation to put your data.
I could babble on endlessly.
One product note - Active passive configurations like Nimble's is one way to do it. However, once you start poking up to 100% utilization, you have to buy another. Have to.
With NetApp we are active active - which means if you start going over 50% utilization on each head, you can still keep on adding workload and do a lot more with the investment. No glass ceiling, and you get to budget an upgrade if you wish on your own terms.
Let me know if you have specific environment or application concerns. I'm happy to help if I can, and if not, I know who can.
2014-03-14 03:40 PM
Hi George -
I currently own both Nimble and Netapp. My experience with Nimble has been great. I agree with Michaels above points of interest when considering a new San technology, so I will speak to them.
1. Support. Since switching to Nimble I have been fortunate to learn what a support experience is "supposed" to be like. When with Netapp, I would often be told "that is a network issue" or "that is a server issue". In some cases they were correct. When calling Nimble support, for one I do not get a help desk receptionist, and I don't have to talk to 5 people before I can get to one that can help me. Every support professional I have spoken with is very knowledgable in not only Nimble support, but also the surrounding products. I have found them to be more helpful in issues dealing with Microsoft and Vmware then the support provided by Microsoft and Vmware.
2. Integration with others. Again a great point. I mostly answered that in topic one, but I will continue. Nimble has proven to me to be just as integrated into key hardware and applications and Netapp was. Their engineers can troubleshoot issuse with other vendors at a much more technical level BEFORE having to call the other vendors to get support from them as well. They are also willing to work with the other vendors support without me having to buy a special suite of hardware and software.
3. Product maturity. Are we talking about product maturity or company maturity? As far as company maturity goes, I understand Nimble hasn't been around nearly as long as Netapp. However, if you do some research you will find that the founders of Nimble are the same engineers and individuals who lead way in the San design over the years. Suresh Vasudevan- Senior executive at Netapp for 10+ years. Varun Mehta- employee #11 at Netapp. Before that he was the 12th employee and served as the vice president of Engineering at Data Domain. Umesh Maheshwari- one of the early architects at Data Domain and helped develop their de-dup file system. So as far as product Maturity you have some of the leaders in storage coming together to make a product that takes advantages of technology that the bigger and more "mature" companies can not, due to legacy systems.
Now for my experience, it was night and day. It started with Netapp being unwilling to give us any credit for the 100's of thousands of dollars we had spent with them in the past. We needed to downsize our physical footprint and increase performance. I could have done that wiht Netapp if I was a customer displacing a competitor san, but since I was an existing customer, my only option was to pay a ridiculous amount. My second option was to wait a year and budget more money to replace the Netapp the following year. The third option was to switch to Nimble for less than $10,000 more then the Maintenance cost of my current Netapp would have been.
As far as performance goes, I have a middle tier Nimble and I have 95% of my data reading from flash. This includes a full VMware server environment, VDI, 4 SQL server, Exchange, and 2 ediscovery applications all reading from the same array. In the event that you do run out of performance, what Michael says is NOT true. You do NOT have to buy another. Nimble offers options of upgrading Processing, Memory, and Flash to increase performance.
All in all I can't sit here and say that my experience with Netapp was bad, but I can say that so far after 2 years with Nimble, I will never go back to the old way.
2014-03-14 05:59 PM
Like Richard S, I too own Nimble, Netapp, as well as an Equal Logic. My Equal Logic replaced the Netapp due to the renewal cost and, at the time, requirements for additional failover controllers.
I've been working with the makers of Nimble since they started Data Domain, recently purchased by Dell a few years ago. Nimble's support is top of my list of support persons which help me out, go the extra mile, and take care of me like I think support should. They really don't seem to care why or who's fault the concern is; they just take care of you.
I would also agree with Michael S, a purchasing is a purchase and sales are sells. When you purchase Nimble, it'll really raises your bar for what you expect your other vendors to live up to.
With all that I've down with Nimble, Netapp, and Equal Logic; my experience with Nimble product, support, renewal: I'll never go back! Sometimes I send Nimble support an email thanking them for the great product and support, apologizing for not calling more often. When I do make a call it's like having additional staff, very responsive, take ownership of the concern, very friendly and polite. It reminds me of the old "Sesame Street" Song Who are your friends in your Neighborhood: it's Nimble!!
2014-03-14 08:57 PM
I agree with pretty much everything said here. We run both and like both but for value, Nimble wins for us. We have a pair of 3240s that worked well but when the load reached a certain level, they just crawled. CPU would be 40% and then jump to 90+ and stay there all day. Latency went triple digits. We had a NetApp engineer onsite and he told us what we needed to do but it meant buying yet another shelf and spending many hours shuffling data around. Oh that vol move? Can't do it live if you use nfs or vfilers. Nimble doesn't support either one of course but we were able to move all our production iscsi loads to a pair of 460s and got better performance and easier management. Both cost us less than what support costs for the 5 filers we have. Disclaimer: my company just bought a Nimble reseller but we bought ours earlier. Our filers are now just used for AIX (Nimble doesn't support it yet) and tier 2 storage.
Speaking of nfs, I've been testing a Linux vm running nfs serving files to our AIX/DB2 systems and so far Linux/Nimble beats the 3240s on every realistic io pattern I try, usually by a large margin. This is with no tuning at all. I've heard good things about running Win2012 as well. Sure, NetApp does CIFS but it's not a full Windows/AD citizen. And don't get me started on the "Non-disruptive Upgrade" with CIFS having to be down the whole time. Our core file share luns are on a filer but with a Windows cluster handling the files.
When we got our first Nimble it was in production within a week. The Nimble SE walked me through the setup and we were online in minutes. The NetApps took a month with an SE's help. Not many dials translates into not much to screw up. I've trained several other admins on provisioning the Nimbles. No way I would turn them loose on a NetApp without a lot of hand holding.
Not multi-tenant ready (no vlans or data segregation outside of initiator groups).
Single login so no roles, etc. These things are being addressed soon I hear.
No AIX - I'm working on that.
Nimble's backup set isn't very strong but I've decided to stay with storage agnostic backup solutions like TSM and vRanger to avoid lock-in. It's not as efficient maybe but I won't sweat the next tech transition. On the plus side, snapshots are stored outside the volumes so no more worrying about luns going offline or file systems getting full just because of snapshot growth.
If you need vfilers or nfs out of the box or can justify the cost, you won't regret getting a NetApp. I've never had trouble with support but I mainly work with my reseller. Love SnapMirror too but Nimble's replication looks pretty good. Just haven't done any production work with it yet.
With the speed of innovation these days, I expect to be replacing our storage again within 3 years. Nimble should be around at least that long so no worries there. If they have what I need then, for the right price, they can keep my business. Still have a bad taste from NetApps' maintenance renewal cost...if I didn't need vfilers I'd be moving off them.
2014-03-15 02:15 PM
I've read your comments with great interest, unfortunately I can't comment on performance since I can't test them myself. That said, bear in mind that the CS460 is a later Gen system than the 3220 which btw was recently revamped.
Now, as far as the simplicity aspect of it. I do believe Nimble has a simple system and so has Tintri and the rest of storage arrays of that range. But there's a reason and we need to take that under consideration. Simplicity is a function of features, functionality, options, 3rd party integrations, data protection and business continuity processes. The fewer of these a storage array has the the simpler it is, the less "knobs" it has.
Furthermore, while cost and speed as always important aspects, Data protection needs to be taken under serious consideration. How are blocks protected, what happens to services when drives fail, what happens during simultaneous failures. At the end of the day, speed doesn't matter if you can't service I/Os.
Redundant parity (R6 or RAID-DP) has been table stakes if leveraging high capacity SATA drives. We learned this using much smaller drives with our Nearstore product many moons ago and the problems only get worst with larger drives. It's a function of fixed bit error rates while the number of bits on the drives increase greatly. That said drive failures are a near mathematical certainty. So, the point I'm making is, during your next testing, regardless of whom the storage vendor is, you should be expanding the testing scope beyond speeds and feeds and include resiliency and fault tolerant tests as well. For example what happens if, while driving I/O to the subsystem, I pull a drive out? What happens if during reconstruction I pull a second drive? Based on data we've collected over many years of studying drive failure events, a double drive failure is more common during Raid reconstructions. What happens if I yank a back-end SAS cable on the primary path? Can the array protect against events that may cause unnecessary failovers? How does the system protects against lost writes given that high capacity drives can't be formatted with 520bps sectors like SAS or older FC drives could? If I scale out my cluster what is the impact of any one of these events on the cluster? At the end of the day, all storage vendors source drives from the same mfgs, however, the protection levels vary wildly at the array software layer and with significant results and business consequences during failures.
Lastly, you made a reference to CIFS and Non-disurptive upgrades. CIFS for as long as it has been in existence has been a stateful protocol. That means that no fileserver, even Windows Clustered NAS servers, can guarantee continuous client connectivity during upgrades or failover events and connections/sessions would need to be re-established. This has changed though with SMB 3.0 with Continuous Availability (CA) shares and the witness protocol which we fully support with Clustered DataOntap.
With CA shares, file handles are persistent, allowing uninterrupted service during brief network outages and storage failovers. When a storage failover occurs, the witness protocol is used to alert clients to proactively move requests to the surviving storage node.
For client side failovers, clients running SMB 3.0 specify an application instance ID when a file is opened. This ID is then maintained by the appropriate nodes on the NetApp cluster for the life of the file handle. If one client fails, the surviving client can use the ID to reclaim access to the file.
2014-03-17 08:29 AM
Nick you are absolutely right about the importance of reliability. NetApp has a well deserved reputation in this area. My company has some 270s and 840s still in production. I'm not knocking the product. I'm just just saying that if you need performance on a tight budget, there are other options. I get calls all the time from HP, IBM and other vendors wanting me to try their kit. Wish I had the time, just for the sake of education... I tried Nimble and it works for me. At the end of the day we alI have to remember that our job is to "Protect and Serve" with both being important. No one should take these 'bet the business' decisions lightly or succumb to brochureware.
I stand by saying that NDU is not true for 7-mode. Sure, CIFS is stateful. But I can failover my Veritas cluster to upgrade nodes with a lot less down time. Most clients don't even notice. NA requires CIFS to be turned off. This is not really NA's 'fault' because of protocol limitations but it should definitely be considered when choosing to use that protocol on a filer. Haven't looked at Cluster ONTAP yet. In fact, I have a meeting with my vendor and NA SE this afternoon to discuss it. We'll see.
2014-03-17 08:59 AM
In my company, we've been through the gamut with different vendors of storage. Many of the trips have not been pretty as we have grown and acquired and moved to integrate platforms from the various acquisitions. iScsi could have been an issue going to the Nimble platform, but it is not. The capability of the Nimble to support multi-initiator sessions with VMware and Windows file servers provides speed, reliability, and function that doesn't stop.
One of our challenges was to replicate data and make it instantly (or close to) available in the event of failure of a site or system. Nimble excels in this without breaking the bank. Our DR recovery time for full function is tested at less than 4 hours for all systems, and 2 hours for systems on the Nimble. A similar test on Net!pp took nearly 24 hours. That was no contest.
Support is good, upgrades are reasonable when the time comes, and a Nimble array can be functional within a half hour of rack. NetApp (we have 4 systems) has consistently taken 2-3 days of time with one of our engineers and on of Netapps in sessions to fully configure the box. I don't have the time or the staff to devote to that.
For our purposes, we're removing all EMC equipment we had and moving our systems to Nimble. Our NetApps will continue to run while the applications that live on them are in use, but will likely also go away as we continue to grow and expand.
2014-03-18 11:04 AM
I think you'll find the recently published Gartner report (Mar. 7, 2014) useful as it compares/contrasts several midrange systems across several important areas. It's a detailed analysis but worth reading.