- casino online slots

Celerra SAN?

by Administrator on February 13, 2009

I got a call from an integrator late yesterday who wanted some input on EMC Celerra.  I gave him what little I know then asked a bit more about the customer requirements he was seeking to fill.

He said the customer, a healthcare provider, was making its first foray into a “SAN” — or what goes by that name in the storage marketecture of today.  The shop had bought a blade system to process some critical data — about 8TB or so initially and expected to grow by leaps and bounds.  They wanted performance and scalability, and manageability was important too, given the smallish size of their IT shop.  HP was the blade provider, but the reseller who sold it to them (not my guy) screwed the pooch on his pricing for a complimentary HP storage solution.  (No surprise.)  Now, EMC was pushing its Celerra solution to the customer via yet another reseller, and the customer was about ready to ink the deal — based on the brand familiarity of EMC, rather than its product capabilities.

The customer has not received reliable performance data on Celerra — mainly because there isn’t much out there.  EMC doesn’t stand for public testing of its kits.

The customer is being told that a Celerra is a SAN.  I always thought of it as a NAS head slapped on top of a Clariion rig that might offer some FC and some SATA drives inside.  Oh, and a boatload of software features that I may or may not need. 

By the way, the EMC NAS products we have tested in the past were pigs, the SMB-oriented crap sold through Dell, but I have had some more positive experiences with Clariion.

The customer is being told that Celerra scales performance and capacity well.  Damned if I know whether this claim is true. 

The customer is being told that Celerra holds its value over time.  Reference was made to NetApp’s tendency to end-of-life its gear after 18 months, while EMC supposedly takes much greater care of its customer’s investments.  I haven’t made the comparison and I wonder if anyone else has.

Finally, the customer is being told that EMC is a much more trustworthy company from the standpoint of its ability to execute over time than is, say, Xiotech, which my friend in the reseller world wants to suggest as an alternative.  In fact, EMC apparently has a whole playbook it uses when Xiotech is competing for “their” business and “their” customer prospect, and they are giving it both barrells to this customer. 

Note that I am not singling out EMC in this regard:  I have seen similar playbooks in just about every storage vendor I have visited.  Unfortunately, sifting the truth from the bullshit consumes more of a consumer’s time than does making an intelligent and thoughtful comparison between products with respect to their fit with specific application and business requirements.

I agreed to chat further with this customer about the considerations that need to be foremost in his strategy for selecting and deploying storage.  In Xiotech’s favor, he can start small and buy storage in a way that captures the underlying falling cost of disk.  Given Xiotech’s web services-based management paradigm and its growing ecosystem of web services-enabled third party software vendors, including great companies like DataCore, he can custom build readily managed infrastructure while containing costs.  Xiotech’s performance numbers have been validated and re-validated and are the best in the business, though I would encourage everyone to try it with their workload before committing to any vendor’s gear.  Finally, I would ask the fellow to consider the broader context of data management.  As a healthcare services company, HIPAA probably applies, which means that he will need a disaster recovery and an archive capability sooner rather than later.  I would prefer to have a flexible model for implementing these services, again controlled via web services, rather than a lock-in model that will go hand in hand with EMC’s inevitable pressure to buy Centera and other hardware/software stovepipes for data protection.

I will report on how effective the truth can be against EMC marketing.  Heck, I would love to test both solutions under his workload in our labs and report the results here, if he will let me.  Unfortunately, having been down that path before, EMC puts a hard stop on public disclosure of the performance of its gear in the warranty and maintenance agreement.  The fellow needs to understand that this is a nifty way that vendors (not just EMC) use to make sure that storage decisions can never be made in a transparent or scientific way.

That alone makes me recommend Xiotech over EMC in this case.  It is irritating to me that he may go with the latter because of FUD they are laying on him about the “questionable future” of “minor or niche players in the storage market like Xiotech” — a quote from the bullshit that the EMC reseller is laying out about his competitor.  When I compare sales acceleration numbers at Xiotech to the low growth numbers from EMC, I think that everyone needs to start reconsidering who will be here in ten years and who won’t.

Returning to my original question:  is Celerra a SAN, even in the beaten-down idiotic definition of what a SAN is today?  Comments welcome.

{ 5 comments… read them below or add one }

strategery February 13, 2009 at 1:59 pm

Celerra is a NAS head on a Clariion backend. I have 2 NS (22 and 502) running a 2 separate sites over FC with HP blades. We’re running a mix of mysql, MSSQL, Exchange, MOSS, etc all on ESX 3.5. They’ve been rock-solid for us (knocking very had on wood).

We use the Celerra (NAS) for CIFS and NFS file sharing, and we connect directly to the Clariion backend to manage the SAN, creating RAID (10, 5, 3, whatever) sets, LUNs, etc.

It’s not a cheap solution, but they’ve served us well, but YMMV. If you have any specific questions, you can contact me directly at bntaylor78 *at* gmail *dot* com.

Administrator February 13, 2009 at 5:09 pm

Thanks strategery.

Here is another response received via email from someone whose company policy prohibits responding to blogs, etc.

“A bunch of years ago (I truly don’t remember when) a Celera showed up on our computer room floor. I never got a straignt story about how it got there, but there it was.

“It sat there for a year doing nothing – microcode problems. It was supposed to hook to a Clariion and this was the first generation of microcode to do this, and it took a year for them to get it working. When the microcode finally got fixed, EMC sent a guy to set it up. He spent 3 days (or was it 4). When he left it still wasn’t working very well.

“My only real memory of the event is how complicated the box was. The Celerra sat on the floor for some time unused, was finally unplugged and later disappeared.

“We just purchased a some NetApp equipment for a Filenet application. We will be running FC LUNs to some unix systems for Oracle and NAS for the Filenet object stores. The LUNs and filestores will be snapped together
to provide a consistant view of the system.

“We looked at EMC Celerra, Netapp and some other solutions. We tried to be open minded about Celerra – it had been a bunch of years since our initial experience. When we got the Gartner info on NAS systems (yea, we have to do that) they make the comment that “continues to receive reports from users that Celerra is relatively difficult to setup and manage when deployed as a gateway”.

“Our technical evaluation was that both NetApp and EMC would do the job, but we thought the NetApp was superior is many ways and would be easier (a big consideration for a small team with lots of stuff to do). We all thought we would end up with EMC – that’s what always happens (they always under cut other vendors on price and we purchase the least cost solution that will technically work).

“Surprise – we purchased the NetApp system! I’m not sure if NetApp was cheaper, or, if it was all the problems we have been having with EMC (lost some snapshot luns on when a clariion processor rebooted, or the mess that ECC v6.1 is).

“My understanding that the newest baby Celerra has internal disk drives (not a Clariion, but it’s own raid of some kind) and is easier to setup and manage.

“The customer you blogged about seems confused on what he wants:
– a NAS – mount CIFS/NFS for his application
– a san as in iscsi luns – good but not real fast
– fibrechannel san with luns – fast but costly

“You talk a lot on your blog about Xiotech. Sometime, you might blog about whether their ISE lives up to their claim that you never have to replace a drive, and, whether you have ever heard whether a ISE has ever suffered a data loss event.”

Thanks for this informative feedback…

strategery February 13, 2009 at 9:54 pm

(Long Post Warning)
The other comment reminded me of my initial experience with the EMC Celerra…

I started the job where my predecessor had already left. So I had this environment where I was totally ignorant, and some people knew some stuff here and there with homebrew, aging equipment. I already knew from my interviews that we were going to grow (and we tripled in size in 2 years), so we already were looking at VMWare to help us scale with limited space/power. Of course I had no experience with EMC, VMWare, FC, iSCSI, etc. This was 2 1/2 years ago. The NS500 was sitting out in the warehouse (it had been there for 4 months, put in production after 8 months). They hadn’t run any power for it. Away we go…

First problem… After we had the L6-30 whips run into our reconfigured server room, we scheduled EMC CEs to come in and fire it up. The guys get there and we (I) realize that the connectors sent with the unit didn’t fit the L6-30s which we were told to install. So I spent a couple hours trying to explain this, to the point that I had to take pictures and show that they didn’t fit.

Got that squared away, and the CEs were then able to get it turned up. Then they sent an engineer to complete the initial configuration. No big deal. I had zero experience, so they showed me some things and I worked my way around the rest. Their support was helpful working with me. At that point we were just using CIFS/NFS file sharing.

Next up (after about 6 months, some pre-sales help and a nice wad of cash later) they send the same engineer to upgrade the NS502 to an NS502G. So the Celerra effectively becomes a Celerra + Clariion. We went with Cisco (MDS) fiber switches, as the IOS was what I was more familiar with. We schedule the upgrade and downtime. The engineer arrives and asks where this part was (value @50k – basically the SPs had to be replaced to allow the fiber connectivity) that hadn’t come up in any of the previous conversations with the pre-sales engineers. “What part?” So we get this part sent and reschedule the upgrade. The only time we could get fit in was the night before my family left for a 13 hour road trip.

2 guys came out to do the upgrade, but they’d never worked on a Cisco fiber switch either. The upgrade process took a long time, then we couldn’t figure out why nothing was coming up on the switch. After sacrificing some fiber cables and getting really crunched for time, I figured out that we needed to enable the ports on the switch. I spent much of my vacation remotely configuring the zones, disks, and ESX hosts as we were really getting pressed for (virtual) servers. Their SAN Management course I took was helpful, though it was more geared towards the Symmetrix line (versus Clariion).

Other than power outages in our building (we’ve since moved it into a hosting facility), we’ve had no major issues. It’s been great for us. We’ve had our implementation hiccups, but EMC has served us well overall. We’re a small organization, so we decided to stick with what we knew and was familiar with and added an NS22 at another site. We later added FC, and it was upgraded without a hitch.

So now we have 2 sites with the Celerra. The NS22 is a big improvement in getting it setup and configuration over the NS500. They’ve served us well, but I don’t know that NetApp, Equallogic, etc. wouldn’t have served us just as well.

phuml February 17, 2009 at 10:29 am

The Celerra is, and always has been a NAS head.

I have a ton of other questions regarding the why, but I’m sure that the customer’s strategic plans have probably taken care of that.

Administrator February 17, 2009 at 1:10 pm

Had a nice chat with the customer yesterday. His head is now officially spinning as he considers my parameters on effective storage architecture.

First, decide how will you manage your infrastructure now and in the future. Selecting a management approach/product/solution will make a big difference in ownership costs and labor costs downstream.

Second, assess real cost of ownership, not just acquisition cost. If the product you are buying will feature 168GB flash drives with MLC flash RAM that only takes 500K writes before cells die (memory wear), then aren’t you looking at replacing memory modules at a predictable rate based on measurable write activity? Why isn’t the $30K per pop cost of the replacement drive (usually replaced in pairs) not being factored into TCO?

Third, look at things systemically. (I know, it causes autism.) Data won’t always reside on your most expensive spinning rust, unless you want to bankrupt yourself. Why not plan for archiving and data protection requirements when you are building storage?

Finally, corporate financial health is a poor indicator these days. EMC isn’t making any earnings projections. NetApp is laying folks off. Xiotech is, by contrast, showing phenomenal revenues growth for a recessionary economy. Who’s really the marginalized player?

We have to consider costs and performance: both are connected to storage architecture and management.

Previous post:

Next post: