I would like to know the comparison features between VVR and SRDF on Sun Solaris. Could you give me advise on these products?Regards.
VVR is from Veritas and SRDF is from EMC. SRDF is hardware based and runs in the firmware of the SYM array whereas VVR is host based on the server connected to the array.
VVR can do SYNC and ASYNC remote copy. The SYNC method copies all writes to the remote client with an I/O complete only when the data is received at the remote host. ASYNC VVR provides an I/O complete to the host when data is written locally. VVR provides for transaction consistent data at the remote site under ASYN by sequencing each write at the primary site, and writing the data at the remote site in the same sequence as the primary. ASYNC can be used for great distance over IP and SYNC should be used for shorter distances (due to latency over distance) SRDF provides SYNC copy, Semi-Sync copy, Multi-Hop and Adaptive copy. SYNC on SRDF is similar to everyone else's SYNC where data is guaranteed at the remote site do to an I/O complete only when the remote storage array acknowledges the copy. Semi-Sync allows for one write behind, which eases latency a bit. Mult-Hop is a method of using multiple storage arrays in the link to reduce latency problems over distance. Adaptive is a method of breaking off a BCV (business continuance volume) and using that as an image to copy to the remote site. You can schedule how often to do this and the bandwidth you provide will dictate how long it takes to do the copy. SRDF usually requires an ESCON extender from CNT, INRANGE, etc. to provide connections from the storage array to the remote site over your intersite link.
VVR can use a standard NIC to copy the data over IP, but may impact host CPU cycles to do the replication. You need a minimum of two software licenses for SRDF, and VVR requires a License for each host involved in replicating data.
You will need to consider a couple of things:
- R/W ratio of application (more writes means more bandwidth needed)
- Application performance
- Host type, CPU speed, memory, network
- Total amount of data to be copied (you need to provide enough bandwidth for intial copy)
- Network bandwidth
- How out of SYNC you can be
- How long you can afford to be down
- Availability of staff and servers at the remote site in case of disaster
In other words you should make the effort to create a complete DR plan. People and skills are an important aspect of this as was learned on Sept 11. You should train your operations staff at both locations on how to implement the plan, with step-by-step details of the complete recovery process.
Editor's note: Also, take a look at searchStorage's bookstore for more books on storage networking.
This was first published in March 2002