07-18-2016 05:11 AM
I have an query
When we say that the speed of a SAN switch is 16GBps, which is nothing but its bandwidth.
How do we find the actual throughput ?
Iam asking this question, because iam going to do a data migration over SAN and have to come at a conclusion of a rough estimate for transfering around 74TB of data over SAN (Storage to Storage Migratio over SAN)
07-18-2016 01:52 PM
This is a question for the ages: 'How much data can my Brocade switch/fabric process?' If I could tell you the answer, I would surely be rich beyond measure, and be the envy of all network admins throughout the land. Here are just a few of the elements that determine actual data throughput:
1. HBA vendor,
2. SCSI settings in HBA for queue depth, and # of queues.
3. Distance to ingress port on the Brocade switch.
4. type of Brocade switch/director/backbone.
5. Port-group to port-group egress from the switch/director/backbone
6. Buffer credit allocated.
7. Distance from the egress port to the storage device(s).
8. Exchange routing policy.
9. Security and encryption settings if any.
10. Typical frame size(full frame unlikely).
11. Data type, i.e. is it a group of image or video rendering, or is it DB transactional type data.
12. Storage device(s) cache size.
13. Storage device(s) cache allocation R/W, and is it dynamic. Source should be full reads, destination should be full writes.
14. Storage device(s) RAID configuration (read, modify, calc parity, write, flush penalty)
15. Storage device(s) media type(SSD, rotating disk, disk speed, disk cache buffer, etc)
16. Read ahead options on host and write behind options on storage.
17. Source data fragmentation.
18. Copy algorithm.
This is a partial list. Having said that, within the realm of the Brocade fabric there are tools available to you to help optimize. First, you can use the D port mode in a fabric to determine some basics. If you are going from one port to another on the same switch, of course, there will be no need for a D port, however - the setup and testing of ports in D port mode is covered in the Admin Guide which can be downloaded from the my.brocade.com website by becoming a member(free) and going to the downloads, all operating systems, FOS, select your version, then find the Admin Guide.
If you are going from port to port on the same switch, and the switch is a base or rack model, insure that you have the host, and both storage devices on the same port-group. This can be identified on most switches and directors by choosing the ports where the port number on the chassis has a background the same as your host and storage. It will either be silver chassis colored or gray stencil indicating the port group(ASIC) involved. Port-to-port exchanges within the same ASIC are lightning fast because Brocade implements it's patented cut-through-routing and not the store and forward in the other vendors. The front of the frame/exchange is leaving the ingress port before the entire exchange is processed into the ASIC.
If you are going between switches, in fabric mode, this gets a lot more complex. there are many things to tune from the host, to both storage devices, and the ability to manage the gains or losses by a change is also complex. In the same Admin Guide you will find a section on 'Monitoring Fabric Performance'. Select that topic and read up on 'End-to-End Performance monitoring'.
Best of luck, if you provide more fabric info, I can provide more detailed advice. A useful baseline of your fabric can be had by downloading and running a SAN Health check, which is avail for free from the home page of the my.brocade.com location.