EMC, OMG why did we?

My company recently merged with another company, doubling the size and doubling the IT requirement. This has meant that our existing VMware platform with iSCSI EqualLogic was not really up to the task as it was beginning to creak anyway and doubling it to support more than 500-750 servers was unlikely to expand.

Time to look at some grown up storage again.  Usual suspects, EMC, NetApp, Dell and HP.

Key features required were expandability, replication, off host backups at remote site. There were a number of other minor things like making copies of live SQL databases between live and dev systems and storage tiering.

Dell Compellant, HP EVA/3Par, EMC VNX, NetApp were all in the frame.

Hardware wise all came in around the same prices, software wise it was different story for the replication and backup piece, the dell offering went off down the NetBackup route which is stupidly expensive and cost more than the hardware so that was first to go. HP went before we started as dealing with HP is difficult as they tend to turn up mob handed and most of the people aren’t even sure why they are there it gets very messy and disjointed. EMC offered VNX, a Data Domain and the ability to use our existing Backup Exec licencing, and Netapp offered Netapp and Backup exec.

After lots of meetings we ended up in the EMC camp. After some more shenanigans we found out VNX2 was imminent which EMC forgot to mention so we had to wait for that. Up to now we were dealing direct with EMC so needed to engage with a reseller which EMC provided.

This was all to be connected over FCoE at the primary site and iSCSI at the remote site due to costs.

This is where it all started going a bit wrong, after every visit we found out more and more limitation of the EMC solution, first replication is not as slick or as simple as the EqualLogic, the interface is at best clunky and slow, Unisphere is a bit of a misnomer, there is link to Replication but to do anything you have to launch a separate web app, Unisphere only supports storage nothing else and to be honest if you want to do anything it easier to do it from the command line, I create all the scripts in Excel using the concatenate function. There are a number of rules about how you present storage to  servers that seem to be artificial, you create a storage group, you can then add servers and volumes in to that group so they can all talk to each other, whilst you put a volume in multiple Storage Groups, each server can only go in one group, this means that if you want to present a volume to only one server such as log drive, the each server needs to have its own storage group and all the shared volumes need to be in all the volumes which can be messy and more difficult to manage than it needs to be. Simply allowing a server to be in more than one storage group would make things hugely easier to manage and I can see no reason for the restriction. VNX seems to be designed for old school physical servers and their disks, preferable UNIX boxes with a few sops to the modern VMware world.

SQL replication is complicated and imposes some major restrictions on how and where you place the server on the storage to get the application consistent replicas, so much so we are looking a a seperate product to mange all our SQL servers. All replication is crash consistent.

Backup Exec does not play well with EMC. There is no real integration with anything and scripting all the mounting of volumes and servers at the remote site is not a simple thing to do. 

The iSCSI implementation is verging on bonkers, you have to put each port on each controller in a separate subnet, I kid you not, this means that each server that has two nics can only talk to two ports on each controller, I can see no reason for this, the EqualLogics has no issues with all the ports being on the same subnet, maybe something to do with multipathing but why do they make things more complicated than they need to be, the Data Domain does not need this, it also makes it more difficult to use the any kind of storage tiering with other vendors’ products.

However, it is so much quicker than the EqualLogics with much less latency. At least one good thing then.

All in all, EMC is a hardware company that writes basic software to manage their kit. The VNX is still really just a Clariion in new clothes and their software is at best poor, I look at what I can do on the EqualLogic with fond memories despite the many woes and general slowness of the platform, sometimes manageability is king. 

Netapp sells itself as a software company that does storage and to be honest there has been many moments when I wished we had gone down that route instead, if only our account manager hadn’t been such a Muppet. However I am sure that we would have several problems with them as well and EMC is where we are at and the next two months will be spent trying to get the replication, DR and backup pieces working consistently and simply in a usable and manageable way.