Hitachi Dynamic Link
Manager (HDLM)
What is the HDLM: it is a server –based software
solution that directly addresses challenges associated with the single point of
failure.
iSCSI
devices , Hitachi storage system command devices, such as Hitachi RAID Manager
command devices, EMC DMX series, EMC CX series, and HP EVA series are managed
by HDLM and tape devices and the disks on the host devices are not managed by
HDLM.
HDLM features.
1. Multipathing: Multiple paths can also
be used to share I/O workloads and improves performance.
2. Path failover: By removing the threat
of I/O bottlenecks, HDLM protect your data paths and increases performance and
reliability.
3. Failback: By recovering a failed path
and placing it back online when it becomes available, the maximum number of
paths available for load balancing and failover is assured.
4. Load Balancing: By allocating I/O
request across all paths, load balancing ensures continuous operations at optimum
performances levels, along with improved system and application performance.
Several load balancing policies are supported.
Since HDLM automatically perform path health checking, the need to
perform repeated manual path status checks is eliminated.
With multi-pathing, a failure with one or more components still allows
applications to access their data. In addition to providing fault tolerance,
multi-pathing also serves to redistribute the read/write load among multiple
paths between the server and storage, helping to remove bottlenecks and balance
workloads. In addition, distributing data access across all the available paths
increases performance, allowing more application to be run and more work to be
performed in a shorter period of time.
How HDLM works:
1. HDLM driver interfaces with HBA
driver or multipathing framework provided by OS.
2. Assign a unique identifier to paths
between each storage devices and host.
3. Distributes application I/O across
each path according to fail-over and load balancing.
4. When a path fails, all outstanding
and subsequent I/O request shift automatically and transparently from failed or
down path to alternative paths.
Two types of failover happen: Automatic and manual.
Failover keeps your
mission critical operation running without interruptions, storage assets is
maximized, and business operations remain online.
The path can go offline due to the
following reasons.
1. An error occurred on the path.
2. A user intentionally placed the path
offline by using the path management window in the HDLM GUI.
3. A user executed the HDLM commands
offline operations.
4. Hardware, such as cables or HBAs, has
been removed.
You can manually place a path online
or offline by doing the following:
1. Use the HDLM GUI path Management
window.
2. Execute the “dlnkmgr” command’s
online or offline operation.
The algorithms
used for the load balancing are the “round robin”. This algorithm simply
distributes I/O by alternating request across all available data paths. Some
multipath solutions, such as the IBM MPIO default PCM, only provide this type
of load balancing.
If we use
the Extended Round-robin for the load balancing then it distributes I/O to
paths depending on whether the I/O involves sequential or random access:
. For
sequential access, a certain number of I/O is issued to one path in succession.
The next path is chosen according to the round robin algorithm.
. For random
access, I/O is distributed to multiple paths according to the round-robin
algorithm.
To centrally manage the multiple HDLM
instances by Hitachi Global Link manager (HGLM).
By HGLM we
can centrally administrator multiple HDLM multipath environment from single
point of control and consolidate and present complex multipath configuration
information on simplified host and storage centric views.
Summary:
1. Provides a centralized facility for
managing path failover, automatic failback, and selection of I/O balancing
techniques thorugh integration with Hitachi Global link Manager.
2. Eases installation and use through
auto-discovery function, which automatically detects all available paths for
failover and load balancing.
3. Provides one path-management tool for
all operating systems, Includes the CLI that allows administrators flexibility
in managing paths across networks.
4. Provides manual and automatic failover
and failback support.
5. Monitors status of online paths
through a health –check facility at customer-specified intervals, and places a
failed path offline when an error is detected.
