Individual disks are only visible to a certain extent on the current storage systems. BS2000 sees “logical volumes” which are assigned to the LUNs (Logical Unit Numbers). The assignment of LUNs to physical disks is largely hidden for BS2000. The storage systems ensure efficient operation by means of RAID levels.
“Large” logical volumes can also be used. These are volumes which are more than 32 GB in size, see the manual “Files and Volumes greater than 32 GB” [3 (Related publications)].
Logical volumes and physical disks
From the BS2000 viewpoint logical volumes are independent units on which I/O operations can be initiated in parallel. When the current high-capacity disks are used, multiple logical volumes are normally located on a disk drive.
When I/O operations to /from logical volumes which are located on the same disk drive are initiated in parallel, the hardware service time is inevitably extended. Whether the extension of the hardware service time leads to an appreciable performance loss depends on the cache hit rate and the utilization of the disk drive – possibly also by users outside BS2000.
In the case of physical disks which are shared by multiple programs, the times of the I/O requests are not generally evenly distributed. With heavier total workloads, this leads to intolerably long wait periods in front of the disks, in keeping with the laws of queue formation. In other words, the software service time per I/O operation is considerably longer than the hardware service time. In order to minimize these wait periods, the workload on shared disks should not exceed 30%. This demand is based on the assumption that the wait time in front of the disk should never be higher than a third of the hardware service time.
Requirements for configuring storage systems
From the BS2000 viewpoint the disk peripherals can be configured in accordance with various considerations:
high capacity (static storage space in the storage system, in GBytes)
high throughput (in Mbyte/s)
short I/O times (in ms per I/O operation)
As BS2000 has no knowledge of the assignment of logical volumes to physical disks, certain performance requirements must be taken into consideration when the storage system is generated. Specifically, these are the
volume size
number of I/O paths
RAID level
type of server load
(OLTP mode, throughput-oriented batch mode - the main application must be borne in mind)I/O load of the particular volume
(I/O operations per second or general statement: low / high - the peak load must be borne in mind)
In many cases specific configuration of the storage system can or must be left to Customer Support. In particular it must be ensured that multiple logical volumes with a heavy load are not placed on the same physical disk. In the case of mixed configurations with other systems, the special characteristics of BS2000 operation should be pointed out (e.g. very high and very even load which also requires the peripherals to perform reliably).
Data format
You are recommended to use NK data format. Compared to K data format, in which the key information must also included, NK data format makes better use of the disk capacity and the cache memory, thus achieving shorter hardware service times.
Use of SHC-OSD
You can ascertain the following with SHC-OSD:
Which logical volumes are configured on a disk (not in the case of thin provisioning)
The physical disk(s) on which a BS2000 volume resides (not in the case of thin provisioning)
For ETERNUS DX/AF: the RAID group in which a BS2000 volume resides
For ETERNUS DX/AF: which other volumes reside in the same RAID group
The drive workload on ETERNUS DX/AF cannot be determined using BS2000 resources and can therefore not be measured using openSM2. It consists of the sum of the I/O loads of the relevant logical volumes.