I was talking to a friend last week who asked me two provocative yet valid questions, and I thought It would be worth posting about because at the end of our conversation he had opened his mind to a few concepts he had not considered.
Question #1 “Whats all this Unified Storage Hooey about?”
Probably a good place to start for people who are not overly familiar with the term “Unified Storage” is to define it. In my own words “Unified Storage” is about being able to present and manage storage using multiple protocols e.g. BLOCK (FC), NFS, iSCSI and CIFS from the same storage array.
It’s no secret that most of us (myself included) have adopted the mind-set “Virtualize Everything” and with that I don’t need to explain where BLOCK, NFS and iSCSI have a place in the virtualization world as they are all widely used and proven.
You might say that typically people tend to pick one protocol and stick with it and you wouldn’t be wrong, but think about being able to tier storage by protocol rather than how you normally associate tiering in the storage world which tends to centered around disk type e.g EFD, FC, SATA. If you look at the cost of FC switch ports and HBA’s compared to Ethernet ports and adapters, it makes total sence that you might consider putting tier 1 production systems on FC storage while placing your test and development systems on IP storage such as NFS or iSCSI.
Question #2 “What good is CIFS to me in a predominantly Virtual world”
When virtualization started taking off years ago we all went mad, it was all about server consolidation and In most cases we virtualized anything that responded to a ping request. What happened often was people virtualized file servers and ended up with virtual machines with huge .vmdk or rdm’s which generally did two things.
- Used huge amounts of valuable storage
- Caused huge bottle necks and backup windows blew out
So I’ll address those two points and try to also cover some other key points which in my mind, make putting file data on the Celerra a no brainer.
De-Duplication: This feature is actually single instance plus compression. People are always sceptical about the figures which storage vendors put out in the market as they are often not achievable, but I can tell you that I have personally seen 1TB file systems reduce by 50% and in some cases 60% with Celerra De-Duplication enabled.
NDMP Backup: File systems containing user data tend to be quite dense, typically this causes traditional backup systems grief and a full backup window can end up being 4-5 days with large amounts of data. NDMP backups using EMC NetWorker can be configured to pull the data over the network or directly to a fibre connected device (tape or VTL drive) for LAN free backup. Celerra NVB (block based backup) handles dense file systems really well, Ive seen backup windows shrink from days to hours.
Snapshots: If you’ve ever been a backup administrator for a large organization you’ll know that on a bad week you can end up spending way to much time recovering user data. Celerra allows you to configure scheduled snapshots of file systems which are then available via the windows previous version tab built into XP service pack 3 and above operating systems.
Archiving: How much of that user data has actually been accessed in the last 6 months ? Introduce EMC’s File Management Appliance (Rainfinity) and you have the ability to move data from expensive FC disk to cheaper SATA storage while being transparent to users accessing the data.
I could go on and on about the other features such as High Availability, Replication, File Level retention, File Filtering, etc, etc but I think this post might end being too much of a mammoth read so I might pick some of these and elaborate more in up coming posts.
I’m not saying it’s a mistake to virtualize file servers because in certain cases there are advantages e.g. being able to replicate and protect using VMware Site Recovery Manager is a good example where its advantageous, but I am saying the larger the amount of data, the more beneficial it is to consider moving to the Celerra.