Вырезка из документа, на который я указал выше:
Цитата:
All in all, the new cache flush mechanism consists of:
• Every session records updates/inserts to a private list of tables updated
• When a session commits, its session private list is merged into an AOS global list of tables updated.
• At an interval set by the user, the AOS global list is written to the database, in a special row in SysLastValue table.
• At a user configurable interval, querying against cached tables will read the database stored list. If the list from the database has a higher cache version for the queried table, the local cache is flushed. This is on a table basis, so only if the local cache for this one table was used, and if the table was actually updated, will the cache be flushed prior to the read.
• All Thin clients do NOT read the database flush list; instead they query the AOS for its global list of updated tables. This ensures that a Thin Client will not conclude that a cache should flush at another frequency than the AOS does. If the client reads the information from the database it would occasionally read the flush info slightly before the AOS did. This caused the Client to flush the local cache before the AOS refreshed its cache, and so the client would read stale data from the AOS cache, but would have the new Cache Version registered, and so would be one version ‘behind’.
• The above AOS mechanism applies for all Fat Clients as well.
• The above Flush and Test interval is derived from a single MaxCacheSync time parameter and are both half of the MaxCacheSync time.
The net effect of the above is:
• All caches, be it on AOS, Fat Client, or Thin Client will be current within the MaxCacheSync time.
• Inactive clients (or AOS’s) will not cause extra network traffic (that is after half the MaxCacheSync has expired, no additional network traffic will occur).
• Due to the asynchronous write of flush info, network traffic is minimized, even though a lot more tables are now included in the flush report.
После этого, так как у вас нет KR1, внимательно перечитайте сообщение по ссылке, указанной выше