EMC perfumes the pig

by Robin Harris on Friday, 10 June, 2016

I feel sorry for EMC’s marketers: they have to make 10-20 year old technology seem au courant. It’s an uphill battle, but that’s why they get the big bucks.

The latest effort to perfume the pig – hold still, dammit! – is EMC Unity. In a piece that – and this is a sincere compliment – rivals the best writing of “wealth creation systems” websites and late night infomercials, EMC’s Chad Sakac, goes on and on about Unity, the VNX successor.

VNX is huge for EMC
Tucci is trying to keep the wheels on until the Dell acquisition closes. Sakac makes it clear that Unity is critical to the effort:

EMC VNX has an installed base that is well, well north of 100,000 deployed arrays. It has been the most successful EMC platform to date in terms of the number of people and customers, and VNX brings more net new customers to EMC than anything else we do – even as it became increasingly aged.

Bolding added.

Mr. Sakac’s piece makes other interesting points.

  • There’s a lot of new code in Unity, and as he rightly points out, it takes a long time to harden code, even with a staged introduction of several pieces. In my experience the hardest bugs are the interactions between modules, not within modules.
  • There’s no PCIe/NVMe interconnect support. That’s the future, not FC or SAS.
  • This is old-school scale-up storage: forklifts forever!

No mention of RAID HDD rebuild times, but assume days for the latest large capacity drives. Naive buyers won’t know what that means – which may explain who is buying these arrays.

The StorageMojo take
The bottom line here is that even the re-engineered VNX/Unity is aimed at customers who are still heavily invested in legacy technology – or who don’t know what that means. EMC’s – and every legacy array vendor’s problem – is that cheap flash IOPS has destroyed the value of years of array controller optimizations for hard drives.

I divide storage arrays into two groups: legacy, developed before 2000; and modern, developed since 2000. As Mr. Sakac makes clear, EMC remains heavily invested in – and dependent upon – legacy architectures.

External storage isn’t going away, but major vendors can no longer ignore the fact that the most important storage is often internal, where bandwidth is cheap and latency lower, such as in-memory databases. That’s why Tucci is selling EMC to Dell – and why Dell is paying way too much for rapidly depreciating assets.

Courteous comments welcome, of course.

{ 7 comments… read them below or add one }

Petros Koutoupis June 10, 2016 at 9:30 am

You are absolutely correct. Even more so with shift to newer topologies (especially for more performant technologies).

>>No mention of RAID HDD rebuild times,
>>but assume days for the latest large capacity drives.

Well, isn’t that always going to be the case? Data needs to get replaced/restored at some point. RAID or just basic block/object replication, even if it is above the drive stack and distributed across the entire cluster. However, some are much more clever than others in minimizing this impact.

Robin Harris June 10, 2016 at 4:17 pm

Petros, most of the object stores and some of the newer file stores – I think Panasas pioneered this approach – spread the data across multiple drives and when there is a drive failure, they copy the data from and to multiple drives to ensure redundancy. Much lower overhead than RAID 5/6 and the “rebuilds” take minutes instead of many hours.

Peter Ram June 13, 2016 at 1:51 am

Spot on Robin, there is so many new technologies that enables the newer bigger drives. Data is growing so the need of bigger drives is inevitable.
Raid 5/6 is still viable in smaller arrays / drives but with the bigger you have to re think it…. Look into technologies like Distributed Raid, GNR, Grid etc. Is it perfect, no, can they cut down the re build time a lot, yes (4-8 times compared to ordinary Raids).

Rob Steele June 13, 2016 at 9:14 am

If customers are still heavily invested in legacy architecture, they will enjoy Unity… however most of the attention and investment at EMC Dell will be towards hyperconverged infrastructure and software moving forward.

Petros Koutoupis June 14, 2016 at 6:35 am

Robin,

SolidFire does something similar but the decrease in rebuild times is attributed to their ability to redistribute the load and evenly spread it across the remaining drives where copies of the original data reside. This will in turn increase load to a single drive by just a bit. Although I believe that you fall into the same trap when you begin to reach full capacity. Going back to SolidFire, they will “rebuild” the data to the remaining drives in the system instead of rebuilding to a spare. That one failed drive would have contained copies of several data sets/objects (some from other drives).

I guess what I am getting at is, most deployments will define a copy or padding threshold. Take the case of Cleversafe, on a 12-wide system, an object may be divided into 8 chunks, with enough padding of 4 additional chunks to ensure fault tolerance. But if one of those drives were to fail, you will want to replicate the object data storage from the failed drive to the remaining drives; maintaining that original padding threshold. The more objects stored, the longer it will take. Reading a 10TB drive or recalculating 10 TB worth of data can be costly.

John June 18, 2016 at 8:57 pm

VSAN also uses the many to many rebuild you describe. I’m not sure who did this first, but XIV was rocking something similar back in the day, and it’s a marvel to behold when you realize your entire cluster is the hot spare and rebuilds happen in 15 minutes. As these systems scale up they actually become more resilient in some ways.

Unity covers mixed data services, it does at least have a robust vvols implementation (VASA is on the controller) it has a 64 bit file system now and some other bells. EMC is positioning the new Isilon project for high performance NAS if I’m not mistaken. Gone is the single threaded raid logic and other crippling limitations the VNX had. While it’s not going to set any records it is an easy to manage (HTML5 and vvols) “good enough” unified platform is my thought on it.

bpolitis@mindspring.com August 2, 2016 at 8:36 am

In regards to the Dell sale. I think the driver for the sale there is VMWARE and VXRAIL.

I had many storage sales teams coming in and spreading FOD on my plans to migrate to VSAN. Now I’m received calls from EMC reps selling the EMC branded vmware VSAN solution. I’m in a mid market company so it i will take awhile to move uphill to those more naive enterprise buyers; but I think that is what Dell is buying. A trusted EMC sales team to eventually spread upwards into those accounts with next gen technologies. EMC sales teams will start to pitch storage on commodity equipment to large enterprise. And no one else currently prices and supports commodity equipment like Dell.

Leave a Comment

Previous post:

Next post: