Re: Taxonomy draft - web replication problems

From: Wojtek Sylwestrzak (W.Sylwestrzak@icm.edu.pl)
Date: Mon Aug 16 1999 - 09:18:40 MDT


my comments to:

5. Inter-Replica Communication

are marked with #:

5.1 Batch Driven Mirror Replication

      Authoritative reference:
             This memo.

      Description:
             In this model, the replica web server to be updated
             initiates communication with a master origin web server.

# shouldn't it be: "replica origin server"
# and: "master origin server" to be consistent with
# the terminology introduced earlier

# also I know examples where the transfer
# is inititated by the master not the replica.

             The communication is established at intervals based upon
             queued transactions which are scheduled for deferred
             processing. The scheduling mechanism policies vary, but
             generally are reoccuring at a specified time. Once
             communication is established, data sets are copied to the
             initiating replica web server.

      Security:
             Relies upon the protocol being used to transfer the data
             set. FTP and RDIST are the most common protocols observed.

# if this is about *web* replication practice
# than my experience is that most popular protocols
# in use would be FTP, rsync and HTTP.
# CVSUP and rdist are also used, but rather seldomly

      Deployment:
             Very common for mirror synchronization in the Internet.

      Submitter:
             Document editors.

 5.2 Demand Driven Mirror Replication

      Authoritative reference:
             This memo.

      Description:
             In this model, the replica web server acquires the content
             as needed due to demand. This is generally done by web
             server accelerators (reverse proxy) operating as origin
             server replicas. When a web client requests a URL that is
             not in the data set or the replica origin server, the
             replica server attempts to acquire it from a master origin
             server and forwarded on to the requesting web client.

# here I'm confused - this is specific to caching,
# not web replication in general. why is it in this section ?

      Security:
             Relies upon the protocol being used to transfer the URLs.
             FTP, Gopher, HTTP and ICP are the most common protocols
             observed.

# i doubt if gopher or ftp is used to transport www objects ?

      Deployment:
             Observed at several large web sites. Extent of usage in
             the Internet is unknown at this time.

      Submitter:
             Document editors.

 5.3 Synchronized Replication

      Authoritative reference:
             This memo. [Ed note; there is no IETF protocol specified at
                         this time. The editors are aware of at least
                         two open source protocols, AFS and CODA, along
                         with one expired IETF draft
                         <draft-leach-cifs-v1-spec-01.txt> and one
                         proprietary protocol Novell NRS; none of which
                         can be considered an authoritative reference]

      Description:
              In this model, the replicated origin servers cooperate
              using synchronized strategies and specialized replica
              protocols to keep the replica data sets coherent.
              Synchronization strategies range from tightly coherent (a
              few minutes) to loosely coherent (a few or more hours).
              Updates occur between replicas based upon the
              synchronization time constraints of the coherency model
              employed and are generally in the form of deltas only.

# isn't it about distributed file systems like DFS, AFS etc ?
# i think it's beyond the scope of this document.
# otherwise we should consider network file system as well
# like cached NFS or WebNSF

--w



This archive was generated by hypermail 2b29 : Thu Nov 18 2004 - 11:21:26 MST