No relevant resource is found in the selected language.

This site uses cookies. By continuing to browse the site you are agreeing to our use of cookies. Read our privacy policy>

Reminder

To have a better experience, please upgrade your IE browser.

upgrade

VTL3500 LSI MegaRAID T3000 Falcon IPStor

Publication Date:  2012-10-16 Views:  14 Downloads:  0
Issue Description
Devices describe:
VTL item deliver in XX site, the user circumstance is VTL3500 (server with raid card, 11*1T SATA hard disk), NBU backup server (windows 2003) and VTL3500 gigabit interface connect directly, AIX Oracle database is the NBU backup client; Set up the storage space, visual tape, tape type as customers’ requirement. The connection type of backup server and VTL3500 is iscsi type.
Phenomenon:
When we doing backup test, backup the local file to VTL3500 by backup server, speed is 1MB/s, the backup speed is not normal.
Alarm Information
na
Handling Process
VTL3500 local raid card is LSI MegaRAID, examine the Logical Dive configuration.
The attribute of logical disk is “Logical Drives Configured”, the order from left to right is: LD(Logical Drive)、RAID、Size、#Stripes(the number of physical disk)、StrpSz(RAID bar size)and DriveState.
Choose Advanced item, check LD parameter (RAID level and size can’t be changed):
StripeSize:slice size;
Write Policy:write policy
Read Policy:read policy
Cache Policy:controller Cache mode
Examine the Write Policy option:
This option can change Cache mode to Write-back and Write-through:
In this situation the default option is Write-through, it influences the speed of backup, after changing to Write-back, and the backup speed is 30MB /s, the speed is normal.
Write-back type is: controller receives all data from cache, then send the finishing signal to host no matter if the data has written to disk. This mode is good but has fatalness. If the controller power outage when data is still in cache, the data will lost in this case. So we advise the users to purchasing some batteries in support.
Write-through type is: after all the data from cache write to disk, the system will send the finishing signal to host. This mode is safe, but capability is poor. And it’s the default cache mode.
Root Cause
Examine the backup server and VTL3500 network interface arrange speed, is gigabit duplex mode;
Because we backup from NBU backup server, backup client is NBU backup server, backup resource is NBU backup server local file, backup the local file to VTL3500 directly, get rid of the bottle-neck between backup server and client.
Examine the configuration of VTL3500, visual tape configuration is IBM3585 tape, tape type is Ultrium 400G 20 disks, this is the entire standard, there is no bottle-neck between agreement and specification.
Examine the license, it accords with the configuration requirement.
Because VTL3500 use the local raid card and local hard disk to provide storage space, there is no external storage equipment, so the problem is the VTL local IO configuration.
Suggestions
When we configure VTL3500, we should notice if the external storage equipment is S2000. If use local raid card storage type, we need to examine the lun configure in raid card, to avoid from VTL read and write capability lower.
This is LSI MegaRAID card Logical Dive configure Advanced option introduce:
1. StripeSize
Slice data in every physical disk will be divided to data sect, in common sense, when the host system read data in order, the larger slice data can provide better read capability. When your host system often deals with the random read requirement, we should choose the smaller slice policy.
2. Write Policy
This option can set cache mode to Write-back and Write-through:
Write-back type is: controller receives all data from cache, then send the finishing signal to host no matter if the data has written to disk. This mode is good but has fatalness. If the controller power outage when data is still in cache, the data will lost in this case. So we advise the users to purchasing some batteries in support.
Write-through type is: after all the data from cache write to disk, the system will send the finishing signal to host. This mode is safe, but capability is poor. And it’s the default cache mode.
3. Read Policy
This option is set for the logical disk if use Read Ahead, the mode has: Normal、Read ahead,Adapter.
Normal: the default read type, it means the controller doesn’t use Read ahead for logical disk to read data;
Read ahead: the controller use Read ahead mode for logical disk to read data;
Adapter: if we access the disk as rule, we should use Read ahead type. If all the read requirement is random, the controller will change the read type to Normal as some arithmetic. Of course the read requirement may be in order.
4. Cache Policy
Cached I/O: all the read requirement are stored in cache.
Direct I/O: all the read requirement are not stored in cache; this mode will not conflict with cache type, all the data will be send to controller and host system. If we access the same data next time, it will read from cache.

We need to notice the configuration parameter and meanings when we deliver in site, and use optimize configuration to meet the customers’ requirement.

END