1C:Enterprise 8 server cluster is a logical term and represents a set of work
processes handling a set of information bases.
In comparison to 1C:Enterprise 8 application server, a server cluster
provides higher reliability, scalability and utilization of hardware resources.
The Main
Features of Server Cluster
A server cluster may run on one or several PCs (working servers).
Each working server may execute one or more work processes handling client
connections within the cluster.
Connection of new client PCs to a cluster's work processes is performed
basing on analysis of long-term statistics for work processes load.
Interaction of cluster's processes with client applications, database
server or with other process is implemented through TCP/IP protocol.
Server cluster processes can be started as application, or as OS service.
Scalability of
Server Cluster
Scalability of the server cluster can be implemented in several ways:
By increasing the number of work processes running at a particular working
server.
By increasing the number of working servers forming a cluster.
Having several work processes allows reducing the load on a separate work
process, on the one hand. On the other hand, this way it utilizes a working
server's hardware resources more effectively.
Besides, running several work processes improves the server's reliability,
due to isolation of groups of clients working with different information bases.
There is a server cluster administration utility implemented as a snap-in for
Microsoft Management Console (MMC). The utility can be used for the following
actions:
Server cluster administration.
Streaming analysis of DBMS deadlocks.
Control for blocking user connections to information base.
Fault Tolerance of Server
Cluster
Fault tolerance feature is provided with three ways:
Cluster redundancy. Several clusters can be grouped, with one cluster is a
master in the group, and other operable ones are slave. Slave clusters only
read working information from the master (serving clients) and check if
there is a master cluster working. If there is no master cluster, then one
of slave cluster declares itself master starts serving client requests.
Working process redundancy. There are three usage options for any working
process: 'Use', 'Use as Reserved' and 'Do not Use'. If some working process
is terminated abnormally, a cluster stars one of inactive processes and
distributes the workload to that process.
Cluster has a client connection log and creates a connection session for
every client. In case of physical connection failure, a cluster waits for
the connection with the user to restore. In most of cases, after connection
restoration the user can continue working from the point where it had
stopped.
Building a
Server Cluster with Various Operating Systems
Server cluster running various
operating systems
All server cluster processes can work under control of Windows® or
Linux® operating systems. Due to interaction over TCP/IP, one server
cluster may include working servers with different operating systems.
However, there are some limitations for working servers running in Linux®
OS:
Such working servers cannot interact with MS SQL Server.
Operation with COM-objects is not supported.
Kerberos protocol is used for server authentication.
Operation with Internet-connection is not available.
