Back to the main page
StorEdge 3510 FC Array or JBOD
I've been working only with SE3510 FC array and JBOD, so here I will not talk about other systems, like SCSI/SATA arrays.
Of course, you want to assign the IP address to the array so you can manage it out-of-band. Use IP from the private network, rather than public one.
The IP setup or change requires controller reset, so you consider this if changing IP on array already in production.
Also consider using Cyclades products so you can access array's console remotely, it makes life much easier.
Of course, the array comes with the firmware, and it can be accessed using old fashion screen. I use VT100 mode.
Note: much more about firmware version 4.2 can be found in Sun doc Part No. 817-3711-17, so I will write here just some basic stuff.
When logging, you may see tons of messages, press Escape to clear one by one, or Ctrl+C to clear all of them. I guess you want to read some of them first before clearing.
See main menu:
New array comes as single logical drive (RAID 0) mapped to LUN 0 with no spare drives. Of course, you don't want this for production, even for playing around and learning.
The SE3510 Array can have max 8 expansion units, so total number of hard disks is 9 x 12 = 108
Max number of logical drives is 32. Max number of partitions per logical drive is also 32.
Note: 250Mbytes on each HDD is reserved for storing controller metadata.
Still confused? Okay.
Logical Drive is an array (RAID 0,1,5) of more physical drives.
Logical Volume is two or more Logical Drives. On FC supports max 32 partitions.
Local Spare Drive serves as stand by for specific Logical Drive. Automatically kicks in and rebuilt if one member of Logical Drive fails.
Global Spare Drive serves as stand by for any Logical Drive in storage.
Channel connects controller and up to 15 devices. Note that logical drive can have physical drives from more channels.
Logical drive or logical volume can be divided to more partitions (max 32).
Each partition maps to LUN and host see it as separate disk.
Controllers
There are single or dual controller array.
Let's say you have 2 controllers. When booting array, controllers auto negotiate and designate one controller as primary, while second one is secondary. Later you configure only primary one and secondary will sync the configuration.
Controllers also monitor each other and will take over work if one fails.
So you want to have 2 controllers instead of one, since you can also enjoy Write-Back Cache feature.
So what's Cache Write Policy?
- Write-Back - controllers get data to write to disk, stores it in memory buffer and immediately tell (lie) the host OS that write operation is completed. And then actually writes data and this of course improves performance.
- Write-Through - controllers write data to disk and then tell host OS truth, that data is written.
Write-Back is default and it will automatically switch over to Write-Through if next occurs: controller, PSU, fan or battery failure
The 3510 can be connected as:
- Point-to-point - only to the FC switch (full duplex communication and allows one ID per channel). This is called Storage Area Network (SAN)
- Loop mode - used with SAN or DAS (direct attached storage) (allows max 8 IDs per channel but only in half-duplex communication). This is default mode.
DAS tips
- Establish 2 connection between each server and array
- Make sure you have multipath s/w on the server, should be there as part of OS, most likely MPxIO
- Map logical drive each server is using to controllers channel that server is connected.
Thanks to Sun Handbook for such a great picture
Back to the main page