|
25.748 bytes |
Product Information |
Document ID: DDSE-44AQ28 |
|
This document is provided to IBM and our Business Partners to help sell and/or service IBM products. It is not to be distributed beyond that audience or used for any other purpose. |
Vinca StandbyServer - Product overview and comparison with Compaq's Online Recovery
Applicable to: World-Wide
Vinca Product Comparisons Under Different OS's
General
The following is a comparison of Vinca's StandbyServer 32 technology running under different network operating systems. StandbyServer 32 interacts with each of the major network OS's (OS/2, Windows NT and Netware) at the same level within each OS. However differences in design and capability between each network OS produces functional differences in the practical uses of Vinca's product between each OS. The differences between the major OS's which affect the use of Vinca's product can be generalized as follows.
Does the Network OS dynamically scan for new devices. Meaning can the OS recognize the addition or subtraction of drives or other devices while running or must the OS be rebooted to recognize the change in drives.
-Netware dynamically scans.
-Windows NT partial dynamic scan.
-Warp Server must reboot to recognize new devices.
Is the network OS divided into a base OS with a network OS/2 added or is it a single inclusive OS. This design has the ability to effect switch-over time with the advantage going to the OS's divided into two parts.
-Netware single inclusive OS
-Windows NT divided base and network OS
-Warp Server divided base and network OS
How are physical or logical disk drives labeled and assigned to the users?
Are they labeled with a specific drive letter or can they be shared without a specific drive letter?
-Netware Specific drive letter
-Windows NT Specific drive letter or shared
-Warp Server Specific drive letter
How is the connection information for each client handled in each OS. Is that connection information stored at the server or at the client. Does the typical client side of OS maintain responsibility for server connection or is the server primarily responsible. (This affects whether or not a client must reboot and reattach to a server after a switch-over has occurred)
-Netware Connection info maintained at the server. Client must reboot and reattach after a switch-over.
-Windows NT Connection information is maintained primarily at the client. No reboot is needed for a client to reattach to a new server.
-Warp Server Connection information is maintained primarily at the client. No reboot is needed for a client to reattach to a new server.
Basic StandbyServer
Netware
Using StandbyServer 32 under Netware offers both flexibility and some limitations compared to it use under other OS's.. The following is a bullet point recap of those differences.
1. Because Netware dynamically scans for new devices during operation the standby machine does not need to be booted prior to booting the primary server. (To avoid an error message it is recommended that the primary server is boot after the standby machine has come up.) This allows the standby machine to be taken of line for maintenance or manual backups and returned to an 'on-line' status without rebooting the primary server. Also recovery from a fail-over to original condition is also simple and straight forward. The original primary server may be repair and brought back on-line as a temporary standby machine. Any new data is then mirrored and when the mirroring is completed the network can be taken down, the rolls of the two machine reversed back to original functions.
2. Because the Netware OS is a single inclusive OS, the entire OS must be loaded each time a server is booted. When a switch-over occurs from a primary server, the standby machine must unload the #run-time# Netware used during normal operation, and reload from DOS the standard Netware license owned by the company. Reloading the entire OS means that Netware may take longer to completely switch from primary server to standby machine.
3. Under Netware drives are #mounted# and assigned a drive letter before they are available to for users to access. When a switch-over occurs, the standby machine must fully mount the mirrored drives and assign them a drive letter. This process may add to the amount of time nesessary for a switch-over to happen under Netware.
4. Connection information is maintained at the server. This means that after a switch-over occurs, clients must reboot, reattach and log back into the server. (The exception to this rule is if the client is a Windows 95 client. Windows 95 has a special feature built in for Netware that allows a client to automatically reattach to a server after a switch-over has occurred.)
Windows NT
1. Windows NT can be forced to scan for new devices but does not do this automatically. This means that during startup a strict boot sequence must be followed. A standby machine can be brought off line for maintenance or to perform a manual backup from that machine. When the standby machine is brought back on line manual process must be run to allow the primary server to recognize and continuing mirroring to the standby machine.
2. Windows NT is divided into a base OS and network services. During normal operation the base OS is running on the standby machine. When a switch-over happens network services are simply added to the base OS thus reducing the amount of time necessary for a switch-over to occur.
3. Windows NT uses a mechanism for sharing and/or labeling disk drives. This means that after a switch-over, drives are available almost immediately.
4. Connection information for Windows NT is maintained primarily at the client. When a primary server fails users will see the server resources disappear. When the standby machine has finished initializing as the new primary server, clients will see the server resources reappear and may continue working without reattaching or logging back inot the new server.
Warp Server
1. Warp is only able to see devices that are present at the time it is booting. For StandbyServer 32 this means that there is a strict boot sequence. Also if the standby machine is taken off line for any reason, the primary server must be rebooted when the standby machine is available again. Recovery from a fail-over also requires that a strict sequence be followed.
2. As with Windows NT the OS is divided into two parts, Warp and the server code. When a fail-over occure the server code is simply started on the standby server making the switch-over happen quicker.
3. Under Warp Server drives are not mounted but they are assigned drive letters for access by clients. Since the drives do not require mounting the switch over time is shortened. However because drive letters are assigned to the drives, care must be take with the drive configuration so the drive letters assignments will remain consistent with the date sets after a switch-over.
4. Connection information for Warp Server is maintained primarily at the client. When a primary server fails users will see the server resources disappear. When the standby machine has finished initializing as the new primary server, clients will see the server resources reappear and may continue working without reattaching or logging back into the new server.
STANDBY SERVER DEFINITIONS
StandbyServer refers to an entire set of products offered by Vinca to enhance data availability. This family of products has the distinct advantage of flexibility and adaptability to a wide range of customer needs. StandbyServer is fully compatible with the current industry standard Intel based PC file servers, both the latest and fastest machines as well as older existing models. This allows customers to take advantage of the newest in PC server technology while continuing to receive value from older existing servers. The following is a quick overview of different StandbyServer configurations.
STANDBY SERVER
This is the term used for the original implementation of Vinca's server mirroring product. In this configuration there is one active server and a passive standby machine. The active server handles all network requests and functions such as file and print services for all users. The standby machine performs the functions monitoring the primary server for activity and receiving all mirrored data from the primary server. In the event of a primary server failure, either from hardware or software problems, the standby machine detects the failure and automatically initializes itself as the primary server. This configuration has the advantage of allowing a customer to match a newer faster server with an older slower server thus preserving the investment of the older machine. During normal operations users experience the benefits of the faster primary server and in since the older standby machine is only receiving mirrored data, in most cases users are not affected by it's slower performance. Should the faster primary server fail and the standby machine take over as the primary server users would notice a speed difference based on the speed of the older machine. But this is a temporary situation corrected as soon as the original primary server can be repaired and brought back on line.
UTILITY SERVER
Utility server is a configuration of StandbyServer where additional functions are run on the standby machine while it is in standby mode. This would include functions such as print services, data base engine services and communication server services. Under a utility server configuration users would still communicate with the primary server for network services and only back ground services would be run on the standby machine. Should the primary server fail, the standby machine would still automatically detect the failure and take over the functions of the primary server in addition to the functions that were being run as a utility server. This configuration has the advantage of being able to gain additional functionality from the standby machine and performs network services load balancing. It does require that consideration be given to performance aspects of the standby machine.
CO-STANDBY
With a co-standby configuration of StandbyServer two servers, both with users attached and acting as primary servers on a network, are able to back each other up. Should one server fail, the other server detects the failure and automatic takes over the functions of the failed server while continuing to perform it's own native functions. Co- StandbyServer has the advantage of gaining maximum use from each server. However careful consideration should be give to performance and data storage aspects of each machine.
FEATURES
This is a partial feature list. Some features may not be fully available under some operating systems, depending on the functionality of the OS.
Two servers act as standby machines for each other
Also can be used in single-standby mode
Automatic fail-over in the event of a server failure
Increases availability of both data and services
Protection for both software and hardware failures
Static load balancing between two servers for all applications
Dynamic load balancing for selected applications
No single point of failure
Clients automatically mapped to surviving server
Local and remote system management, reports, and alerts through NetFinity
Independent, dedicated high-speed link between servers retaining full network bandwidth
Copper, fiber, and WAN capability for server link, providing local, campus, and remote server mirroring solutions
Communicates over industry-standard hardware
Fully compatible with Intel-based servers
Open and flexible configurations-not limited to specific server, drive, or adapter brands
Allows easy incorporation of both internal and external disk subsystems
Certified by major operating system manufacturers
Allows life extension of older server hardware by use as standby machines
Cost effective
Incorporates all the benefits of Vinca's industry-leading StandbyServer
Product extensions available for backup enhancement, remote data vaulting, etc.
POSITIONING INFORMATION:
Server pair could be used as fail-over element in NT/DEC clustering
Can be positioned as first step toward server arrays
Alliances possible with many major industry players (e.g., Oracle, Novell, etc.)
Announcement could encompass LAN Server, OS/2, NetWare and NT, not just NT, making it the first announcement of its type across operating systems
Multiple OS and alliance opportunities could set de-facto standards for industry
Use of NetFinity as common management tool could simplify management across operating systems
Specific NetFinity hooks could be provided for differentiation from other products
COMPARISON with COMPAQ ON-LINE RECOVERY SERVER
IBM/Vinca Co-StandbyServer Compaq On-line Recovery Server
-------------------------- ------------------------------
Multiple operating system support, NT support only
LANServer,OS/2, NetWare, Windows NT
Will support virtually any Announced alliance with Oracle, no
application. Announced alliances support announced for any other
could include a wide set of vendors. application. Applications must be
registered with the recovery
agent software before full
protection is given.
Can be managed through NetFinity Works only through Insight Manager
and/or other management tools.
Multiple, flexible interconnect SCSI interconnects only, limit of
options between servers-virtually 12 ft between servers.
any distance any kind of
interconnect, including WAN
capability for disaster recovery.
Architectural platform for other Fills one and only one function.
added value enhancements such as Point product-not an architecture.
SnapShotServer and data vaulting.
Easy support for a wide variety Support for external ProLiant RAID
internal and external storage storage only, and must include
including RAID. special SCSI switch.
No special configurations on Requires disabling of the array
storage or channel, accelerator, impacting performance
No performance impact. and having a very large impact on
RAID5 operations.
No requirement for extra storage Requires both local storage and
beyond that needed for standard separate mirrored storage for
operation. All data and services protected data. Data and services
can be protected. Can also be on local storage are not protected
configured so that some data is
protected and less valuable data
is not.
No requirement for additional Required multiple controllers-each
controllers beyond those needed for machine will have a minimum of one
standard storage configurations. local controller, one recovery
controller, one primary controller
Open support for multiple hardware Closed system-supports ProLiant
platforms. only.
The two servers can be widely Allows difference between servers
different in configuration, within very narrow spectrum.
different processors, RAM,
host busses, brands, NICs,
disk drives, etc.
Can be used with currently installed Only can upgrade if installed
systems in an upgrade approach systems are Compaq ProLiants.
regardless of installed product.
No single point of failure through Single point of failure in RAID
simple inexpensive mirroring could eliminate data availability
and duplexing. unless extra cost is spent for
mirroring in top of expensive
RAID systems.
Lower cost. Higher cost.
No usage of COM ports. Requires dedication of a COM port
on each server.
Server "heart beats are detected
only through the com port
connection. NIC or other network
link failures cannot be detected
or automatically recovered from.
Easy implementation through standard Complex implementation requires
mirroring requires little additional education on a wide variety of
education. new ideas, including recovery
agents, primary controllers,
secondary controllers, local
controllers, recovery
interconnects, etc.
Could be configured to monitor both Can detect only faults that stop
the 'heartbeat' between servers the 'heartbeat'.
and the network connection, adding to
faults that can be detected.
Checks for failure over two paths, With only one check(heartbeat) for
both the Vinca interconnection and a downed server it is possible to
the network, before automatic initiate a switchover when the
switchover occurs. No chance for only thing that has failed is the
two servers to be vying for work heartbeat hardware or agent. This
or resources. could result in two servers in
conflict for the same load, data
and disk drives.
Faster switchover for any failure Slow switchover since timeout must
from quicker timeouts, multiple fail be at least 30 seconds before any
checks. recovery.
Restoration of the original Restoration of the original
configuration is quick and simple. configuration requires shutdown
of both servers and all storage
devices, and a significant delay
for availability from either
machine.
Fault Tolerant Technology Overview
In the world of high availability solutions there seem to be many choices. These products can be grouped into several general categories which help to identify their functions and limitations. SCSI Switch One group would consist of the SCSI switch based solutions. This solution connects two servers to a single storage device. Only one of the connections to the storage device is active at a time and that connection can be switched between servers should the primary server fail. Twin tailed SCSI storage devices generally fall into this category as well.
This technology has several limitations including;
A single point of failure is introduced in the SCSI switch itself.
A goal of any high availability solution should be to eliminate all single points of failure.
Most SCSI switch solutions use a single presence check or heart beat to determine primary server status. This again is a single point of failure. SCSI is limited by media constraints. SCSI devices must always be in close proximity to each other. This eliminates SCSI
switch technology as an option in data vaulting or off site mirroring solutions.
As SCSI technology advances in speed and through-put, SCSI standards change thereby introducing incompatibility problems. This can be seen in the current difficulties with fast & wide SCSI adapters working with SCSI switch technology. Compaqs' On-line Recovery server is a good example. Enhanced through-put capabilities of the SCSI controllers must be disabled to achieve compatibility with the SCSI switch.
Because SCSI switches are typically hardware only solutions, they offer a lower level of adaptability to the OSs' on the server. This
means that the switch-over process from primary server to secondary machine may involve more processes and/or take longer.
File Copy Products
A second group would be the file based copy products. These are software based solutions which periodically copy files across the
network from a primary server to a target server. These copied files can then be accessed on the target server should the primary server fail. This can be compared to a normal backup process, only instead of backing up to a tape drive, the backup is sent to another server with the data remaining in a more usable format.
The general problems with this technology are as follows;
Copy latency is the primary shortcoming of this technology. As data is written to the primary server it is generally buffered and later sent to the target server. Data sets between the primary server and the target server can be out of sync by as much as twenty minutes. In a high transaction based system this is generally unacceptable.
File copy solutions also tend to use only a single presence check and therefore have a single point of failure.
The typical configuration for file based copy products is to send the file copies across the network to the target server. This can dramatically reduce precious network bandwidth.
To accomplish the file copy process, several of these products install themselves into the network OS in a very awkward and unnatural way. This in turn creates an unstable environment.
Transaction Based Solutions
A third group would be the transaction based products. These are solutions which duplicated data between two servers on a transaction based process rather than a file copy based process. There are a much smaller field of products in this solution category. Vincas' StandbyServer as well as Novells' SFT solutions are a members of this technology category. Vincas' StandbyServer family of products overcome the limitations of SFTIII as well as the other high availability solutions.
Features and benefit include.
No single point of failure.
Mirrored data is kept on a separate data link preserving network bandwidth.
Fully compatible with all utilities and applications run on the primary server.
StandbyServer acts as a simple, natural extension of the network OS. This makes it easy to install and to administer.
Customizable error detection and switch over parameters allowing the system to be tailored to the specific needs of the environment.
Compatible with industry standard hardware.
Does not require identical hardware between mirrored servers.
There is no latency or 'spoofing' in the data transfer channel. This means that data is written to the standby machine at the same time it is being written to the primary server.
Vincas' StandbyServer family of products are the best choice for total server failure protection.
|
Search Keywords |
| |
|
Document Category |
Technical Specifications | |
|
Date Created |
19-01-99 | |
|
Last Updated |
27-01-99 | |
|
Revision Date |
27-01-2000 | |
|
Brand |
IBM PC Server | |
|
Product Family |
Clustering | |
|
Machine Type |
Various | |
|
Model |
| |
|
TypeModel |
| |
|
Retain Tip (if applicable) |
| |
|
Reverse Doclinks |