Introduction
Storage and fault-management are critical components of RAID systems. Improper implementation can result in system fault-tolerance becoming jeopardised. This technical brief discusses Mylex’s Global Array Manager features and its ability to implement Adaptive Storage Management. Adaptive Storage Management is Mylex’s solution for providing enhanced storage management such as fault management, storage management and storage monitoring through remote or local access.What is Global Array Manager?
Global Array Manager™ v2 represents a significant change over previous versions of the Global Array Manager (GAM) software. GAM, as it is now referred to, is a complete and robust client/server storage management application that can be used for monitoring, managing and for configuring Mylex controlled storage either locally or remotely through standard TCP/IP connections.
GAM’s features include array configuring, dynamic performance monitoring, industry standard fault-management, remote TCP/IP support, reliable messaging and superior operating system support.
RAID Configuring Through GAMArray configuring through GAM represents a significant breakthrough for RAID. To this date no other manufacturer of RAID controllers provides a storage management application featuring configuration capabilities in a single, integrated tool set.
GAM’s methodology for RAID configuring is based on providing simplicity to the user. The main focus of GAM’s new configuring capabilities is RAID Assist™.
RAID Assist is the industry’s first true RAID wizard for configuring sophisticated arrays. RAID Assist approaches RAID configuration from the view point that the system administrator might not be familiar with the terminology and strategies associated with RAID. With this approach towards RAID configuration, RAID Assist’s assisted configuration reduces the typical learning curve associated with RAID to a simple four step process with progress tracked on the RAID Assist clipboard.
More advanced configurations can be achieved through RAID Assist’s advanced configuration interface. Even in advanced configuration, RAID Assist simplifies RAID set up with a versatile drag and drop interface. Array optimization can include adjusting stripe size, cache line size, write cache policy and controller clock rate. Settings such as these often require adjustments when multiple drive makers are used in a storage system environment.
RAID Assist is a simple, yet powerful, configuration wizard.
Fault Management Through GAMThe growth in popularity of the Internet as well as the rapid implementation of Intranet servers within companies has forced the industry wide acceptance of TCP/IP as a transport protocol. What this means to fault-management is nothing short of wider network support.
GAM leverages several benefits of TCP/IP that are not present in other RAID tools due to the lack of TCP/IP support. One benefit readily obtained is the ability to support Windows, NetWare and UNIX platforms with a common source. In other words, GAM supports heterogeneous network environments through a single client interface. This eliminates the need for special clients to monitor the Windows NT servers, NetWare servers and UNIX servers.
The Use of TCP/IP also allows GAM to send fault-notifications in a reliable fashion. Management applications which use SNMP or DMI as a sole means of messaging have a fundamental limitation when it comes to sending events. Once an event is sent from the server, no mechanism is available to ensure the event is acknowledged and dealt with.
Reliability is inherent with TCP/IP. Take a simple Ping command to an IP address for instance. When the Ping is sent, a resulting reply is sent from the communicated client if the connection is valid. If the connection is not valid, an indication of "request timed out" is received by the server. In this fashion the GAM equipped server is capable of ensuring that a designated client receives its message. If a reply is not received, the message will continue to be sent.
In addition to reliability, GAM also ensures industry standard fault-management through such enclosure interfaces as AEMI, SAF-TE and SMART. These standards are used by peripheral and system manufacturers to provide failure indications in the event of power supply, fan, or disk drive failure. If any one of these components fails within a server, costly data loss could result. Typically, higher end servers implement redundant power supplies and fans in addition to RAID. The idea behind such a strategy is to minimize any single point of failure within the server. However, lack of enclosure management support in the management software virtually undermines the very concepts of redundancy and fault-tolerance. The thought of a tree falling in a forest with no one around to hear it fall comes to mind. If the management software doesn’t support enclosure management, the system administrator may not learn of a component failure in time to replace it and restore the server to a state of redundancy.
Also in keeping with the subject of fault management is consistency checking. Consistency checking is relatively essential in ensuring that data integrity is maintained within a RAID-5 array. Since RAID-5 makes use of parity information to reproduce potentially lost data during drive failures, it is important for the parity information to be periodically checked against actual data. GAM allows this simple test to be performed in the background by simply double clicking on a logical drive icon and then selecting to execute a consistency check on the RAID-5 volume. Through this test, system administrators can ensure the proper functioning of a server’s storage system prior to a critical failure. In essence, fault-management entails more than just reporting device failures. It includes predictive failure analysis as well as the ability to perform preventative maintenance.
Network Security and GAMNetwork security should not be confused with message reliability or with fault-management. Those issues were discussed earlier in this document. Instead, network security deals with preventing or not allowing any unauthorized access to a server. The Global Array Manager is a powerful tool that has remote configuration capabilities. These capabilities lend themselves to the system administrator in times of crisis when a device has failed resulting in the server’s availability being either threatened or compromised. However, in the hands of an unauthorized person, these tools can be devastating to a network.
Guest Access
GAM has taken this security issue into consideration with its mechanism for
controlling various access levels. There are essentially three access levels implemented
in the GAM architecture. The first, Guest Access, requires no server access privileges at
all. Guest Access allows a user to monitor an unlimited number of attached servers for
fault information. This system for reporting of faults is called GAMEVENT. GAMEVENT allows
an administrator to configure the server with an unlimited number of Client IP addresses
which to send any event notifications. Since the information is sent using TCP packets,
GAMEVENT ensures delivery to the destined clients or re-sends the packets. If the destined
client is running the GAM Client console, the event will be updated in the event log and
the appropriate alarm will be issued. GAM Client can issue alarms through pages, faxes or
e-mail in addition to its graphical representation.
User Access
User Access allows the user to monitor additional parameters such as controller
performance and I/O bottlenecks. These capabilities can become useful when deployed across
a large network to ensure all monitored servers are functioning properly. If it appears
that a bottleneck condition is occurring, the user can then determine what actions ought
to be performed to eliminate the condition before a failure occurs.
Gamroot Access
Gamroot is a superset of typical administrator access. In many instances Gamroot and Administrator will be one in the same. However, in some network environments,
more than one user may be granted Administrator privileges. In this instance, GAM
differentiates the true Administrator by this new designator. Gamroot allows the user to
perform all GAM Client functions. This includes remote RAID configuring, making drives hot
spares, adding additional storage as well as killing drives.
Security controlled at the server
Another security precaution implemented in GAM is to control security at the server
instead of at the Client. Other management programs will typically control security in the
opposite fashion. The problem with this in a large network environment is simple. Not all
companies have just one person maintaining all of the servers. Often times there is an
administrator over each departmental server. With this being the case, it is better not to
allow Gamroot Access to all attached servers by one administrator unless it is
intentional. Therefore, GAM uses the proven security mechanism inherent in the Network
Operating System (NOS). If the administrator is granted User or Administrator privileges
to the server, the NOS will provide the necessary indication to the GAM Client. All
Gamroot functions are enabled in this fashion and can only be implemented through the GAM
Client or through the use of the GAM API.
SNMP and DMI support without GAM Client
Adding storage to a server has, in the past, been a grueling process. Typically it involves either logging users off of the network or waiting till off hours to perform the necessary operations. This is no longer necessary when GAM is used with a Mylex controlled RAID system. All that is necessary is the ability to place additional drives onto an active SCSI bus. This is not a problem when an active back plane is implemented that supports hot-swapping of devices.
Once a drive or a group of drives is added to a server, the user can use GAM to scan the server’s SCSI bus for new devices. This can be performed through the controller options. Once the newly available devices have been recognized by GAM, they can be configured using either the Add Logical Drives or the Advanced Configuration buttons of RAID Assist.