Really not knowing where to being, I'll start with the main storage unit known as the
symmetrix.
Hardware:
The symmetrix being an enterprise class unit manages arrays in a way not typical to a JBOD solution. Internally the unit creates, in a sense, a virtual disk. The basic middle fo the road unit (DMX1000 is by no means cheap, it simple holds fewer disks and cache) is capable of holding 144 3.5" fibre channel hard drives.
These drives are never "seen" by a host as individual drives. Using the internal software, an EMC engineer can create thousands of volumes based on the amount of storage available to the unit. (Users can also create their own volumes through the use of specific EMC software) My specific group has tested storage units containing 4000+ drives. Other groups have test larger amounts. (Virtual, not physical)
When a devices is created, there are a plethora of options available to the user. Being an enterprise storage unit, these arrays are capable of housing CKD devices and FBA devices both at the same time. This allows for mainframe and open systems host connectivity for a mixed environment, all using one array.
I'm more familiar for FBA devices because I typically work in a windows environment. More recently I've been into the Unix environment as well, and they both use the same type of volumes (FBA).
A volume can be created to a specific size based on the user's need and available space. A user can create from a variety of types, such as:
a Standard device
A Meta device (striped/non-striped)
A BCV (business continuance volume)
A SRDF Device (symmetrix remote data facility)
RAID-S (special RAID configuration, not typically used)
(there are a few others that I haven't work with yet, so I'm not mentioning them)
All device types can have up to four mirrored devices added to them. For example, a user can create a Standard volume, and have 4 mirrors added for data protection. Each mirror uses an algorithm to place the data on different physical disks and different data buses to increase reliability, data protection and uptime. One or more of the mirrors can be an SRDF mirror.
SRDF
The unique volume is the SRDF volume type. This type requires that a user has an additional compatible symmetrix unit due to the nature of this volume type. An SRDF volume is essentially a geographically distributed mirror. A typical (but not limited to) use for this volume would be a bank or financial institution. This type of customer would create a volume on one symmetrix unit in a New York datacenter, and have the other symmetrix unit in a New Jersey datacenter. The volumes can be synchronously (or asynchronously) mirrored between the two states to protect the company from a disaster scenario such as the world trade centers.
BCV
BCV volumes are not typically used as direct storage for a host. A BCV (as the name implies) is a continuance, or as I like to call them, check point volumes. These volumes are design as a floating mirror. Through the use of
software, these volumes are used to create a snap shot of your production volume. A typical use for a BCV is in a database environment. A customer may be using an Oracle database, and they can't take the database offline for any length of time because they will lose money due to downtime.
Using a BCV, a customer can schedule the synchronization of this volume with the production volume. Once the BCV has synchronized 100%, the customer will issue a "split/hot split". This will give them a point in time copy of their production database. For an environment, which requires multiple check points, the user can configure numerous BCV's (lets say 12 as an example) and they can have 1 BCV for every 2 hours of the day. This will give them 12 "check points" in time of their production database. Through the use of software, they can schedule the synchronizations at any time they desire.
BCV's have the ability to perform 100% establishment which copies all data tracks from source to target. (Target device being the BCV) Once the first establish has been performed, the user can split the BCV so no more tracks are updated. As the production volume continues to be used, tracks will change and the BCV will know which tracks have changed. So the next time a user performs an establishment, they can run an incremental establishment, which will copy over only the changed tracks. This will save a ton of time for users who have very large production volumes. BCV's can also be mirrored to protect the point in time copies.
If the user's production volume becomes corrupted, they can issue a restore command that will take the contents of the specified BCV and overwrite their production volume.
Host Connectivity
To complicate things even more, each host addressable volume is assigned a LUN to a specific director on the array. Depending on the configuration a unit can have multiple port directors and multiple port/multiple processor directors. These directors contain the fibre channel connections, typically 2 Gb GBICs in a 4port and 8 port configuration. The connections would then typically lead to either the host (for direct attachment) or to a switch/fabric environment.
VCM (volume configuration management, AKA LUN masking)
Through the use of WWN (world wide names), a host can be assigned a range of volumes available to a director. For example, one directory may have 200 devices available on it and that directory may be attached to a switched fabric environment. In this fabric, may be 35 hosts or more. Through the use of software, one host can be assigned 30 out of the 200 devices available on that director. This is done by taking the WWN of the HBA fibre channel adapter and entering this into the VCM database. Essentially the database can be configured to grant access of this specific WWN to a specific range of volumes.
I have to get going for a bit, so I will write some more information later if you are not too bored from the above information.
