EMC Storage Product Knowledge Sharing

We have discussed regarding Symmetric DMX-3. Lets talk about Symmetrix Device. DMX-3 system applies a high degree of virtualization between what host sees and the actual disk drives. This device has logical volume address that the host can address. Let me clear that “A symmetrix Device is not a physical disk.” Before actually hosts see the symmetrix device, you need to define path means mapping the devices to Front-end director and then you need to set FA-PORT attribute for specific Host. Let not discuss configuration details now. I am trying to explain what Symmetrix device is if this is not physical disk and how it will be created.

You can create up to four mirrors for each Symmetrix device. The Mirror positions are designed M1, M2, M3 and M4. When we create a device and specify its configuration type, the Symmetrix system maps the device to one or more complete disks or part of disks known as Hyper Volumes/Hypers. As a rule, a device maps to at least two mirror means hypers on two different disks, to maintain multiple copies of data.

Most of user asked me what are basic differences between EMC Clone/Mirror/Snapshot? This is really confusing terminology because most of things will be same logically.Only thing change that is implementation and purpose of uses. I am trying to write basic and common differences:

1) A clone is a full copy of data in a source LUN. A snapshot is a point-in time "virtual" copy that does not occupy any disk space.
2) A snapshot can be created or destroyed in seconds, unlike a clone or mirror. A clone, for example, can take minutes to hours to create depending on the size of the source LUN.
3) A clone or mirror requires exactly the same amount of disk space as the source LUN. A snapshot cache LUN generally requires approximately 10% to 20% of the source LUN size.
4) A clone is an excellent on-array solution that enables you to recover from a data corruption issue. Mirrors are designed for off-site data recovery.
5) A clone is typically fractured after it is synchronized while a mirror is not fractured but instead is actively and continuously being synchronized to any changes on the source LUN.
Clones and mirrors are inaccessible to the host until they are fractured. Clones can be easily resynchronized in either direction. This capability is not easily implemented with mirrors.
Restoring data after a source LUN failure is instantaneous using clones after a reverse synchronization is initialized. Restore time from a snapshot depends on the time it takes to restore from the network or from a backup tape.
Once a clone is fractured, there is no performance impact (that is, performance is comparable to the performance experienced with a conventional LUN). For snapshots, the performance impact is above average and constant due to copy on first write (COFW).

I left one more term EMC BCV(Business Continuity Volume). It is totally different concept thought. I will try to cover in upcoming post though I have discussed about EMC BCV in my older post. But it is more or less cloning only only implementation change.

As we know that we have different type of RAID but all the raid type are not suitable for the all application. We select raid type depending on the application and IO load/Usages. Actually there are so many factor involved before you select suitable raid type for any application. I am trying to give brief idea in order to select best raid type for any application. You can select raid type depending on your environment.

When to Use RAID 5
RAID 5 is favored for messaging, data mining, medium-performance media serving, and RDBMS implementations in which the DBA is effectively using read-ahead and write-behind. If the host OS and HBA are capable of greater than 64 KB transfers, RAID 5 is a compelling choice.
These application types are ideal for RAID 5:
1) Random workloads with modest IOPS-per-gigabyte requirements
2) High performance random I/O where writes represent 30 percent or less of the workload
3) A DSS database in which access is sequential (performing statistical analysis on sales records)
4) Any RDBMS table space where record size is larger than 64 KB and access is random (personnel records with binary content, such as photographs)
5) RDBMS log activity
6) Messaging applications
7) Video/Media

When to Use RAID 1/0
RAID 1/0 can outperform RAID 5 in workloads that use very small, random, and write-intensive I/O—where more than 30 percent of the workload is random writes. Some examples of random, small I/O workloads are:
1) High-transaction-rate OLTP
2) Large messaging installations
3) Real-time data/brokerage records
4) RDBMS data tables containing small records that are updated frequently (account balances)
5) If random write performance is the paramount concern, RAID 1/0 should be used for these applications.

When to Use RAID 3
RAID 3 is a specialty solution. Only five-disk and nine-disk RAID group sizes are valid for CLARiiON RAID 3. The target profile for RAID 3 is large and/or sequential access.
Since Release 13, RAID 3 LUNs can use write cache. The restrictions previously made for RAID 3—single writer, perfect alignment with the RAID stripe—are no longer necessary, as the write cache will align the data. RAID 3 is now more effective with multiple writing streams, smaller I/O sizes (such as 64 KB) and misaligned data.
RAID 3 is particularly effective with ATA drives, bringing their bandwidth performance up to Fibre Channel levels.

When to Use RAID 1
With the advent of 1+1 RAID 1/0 sets in Release 16, there is no good reason to use RAID 1. RAID 1/0 1+1 sets are expandable, whereas RAID 1 sets are not.