233
Fig. 1. VAX / VMS cluster.
The Computer Interconnect (CI) bus operates at 70 Mbps and is used to
connect VAX computers and HSC controllers using a Star Coupler switch. Each CI
link has double redundant lines, two for transmission and two for reception, using
basic CSMA technology, which uses a specific delay node for resolving collisions.
The maximum length of a CI link is 45 meters. The Star Coupler star switch can
support up to 32 CI buses, each connecting a VAX computer or HSC controller. The
HSC controller is an intelligent device that controls the operation of disk and tape
drives.
VAX computers can also cluster in a local Ethernet network using NI -
Network Interconnect (so-called local VAX clusters), but the performance of such
systems is relatively low due to the need to divide the Ethernet bandwidth between
cluster computers and other network clients.
In early 1992, DEC announced support for building a cluster based on Digital
Storage System Interconnect (DSSI) bus. On the DSSI bus can connect up to four
computers VAX lower and middle class. Each computer can support multiple DSSI
adapters. A separate DSSI bus operates at 4 MB/s (32 Mbit / s) and allows up to 8
devices to be connected. The following device types are supported:
DSSI system
adapter, RF series disk controller and TF series ribbon controller. DSSI limits the
distance between nodes in a cluster to 25 meters.
Worldwide, there were more than 20,000 installations of VAX clusters.
Almost all of them are built using the CI bus interface.
VAX Cluster System Software
To ensure that the processors interact with each other correctly when
accessing shared resources, such as disks, for example, DEC uses the Distributed
234
Lock Manager (DLM) distributed lock manager. A very important function of the
DLM is to ensure the coherent state of the disk caches for Input/Output operations
of the operating system and application programs. For example, in relational DBMS
applications, DLM is responsible for maintaining a consistent state between database
buffers on different computers in a cluster.
The task of maintaining the coherence of the Input/Output cache between
processors in a cluster is similar to the problem of maintaining the coherence of the
cache in a strongly coupled multiprocessor system built on the basis of a certain bus.
Data blocks can simultaneously appear in several caches and,
if one processor
modifies one of these copies, the other existing copies do not reflect the current state
of the data block. The concept of block capture (block ownership) is one of the ways
to manage such situations. Before a block can be modified, possession of the block
must be secured.
Working with DLM has significant overhead. Overhead in a VAX/VMS
environment can be large, requiring the transfer of up to six messages on the CI bus
for a single Input/Output operation. Overhead can be as high as 20% for each
processor in a cluster. Database providers, when using a dual-processor VAX
cluster, usually expect to obtain an increase in throughput of 1.8 times for selection
transactions and 1.3 times for database update transactions.
Conclusion
In place of VAX - clusters came Unix - clusters. At the same time, VAX -
clusters offered a proven set of solutions that sets the criteria
for evaluating such
systems.
Today, Unix clusters are lagging behind VAX clusters in functionality. The
greatest difference is due to the implementation of customer recovery in case of
failures. In VAX clusters, such restoration is performed by the software of the VAX
cluster itself. In Unix clusters, these capabilities are typically implemented by a
separate
level of software, called a transaction monitor.
Do'stlaringiz bilan baham: