ando@OpenLDAP.org wrote:
provide an interface to remove a set of entries from the cache
The rationale behind this is that, for bandwidth occupation issues, we need to cache frequent proxied searches that retrieve very large objects. For this reason, we want relatively long TTL, but we also need to be very tight on updates on the remote server. For this purpose, I'm designing a layer based on the ldap_sync API (RFC 4533) that performs a persistent search looking for changes on the modifyTimestamp of the sensitive objects (small data that always gets modified, to be conservative) and, in case, removes those objects from the cache in a consistent manner. If this feature works as intended, and is considered useful, I can commit it (maybe it could become an integral part of pcache...).
p.
Ing. Pierangelo Masarati OpenLDAP Core Team
SysNet s.r.l. via Dossi, 8 - 27100 Pavia - ITALIA http://www.sys-net.it --------------------------------------- Office: +39 02 23998309 Mobile: +39 333 4963172 Email: pierangelo.masarati@sys-net.it ---------------------------------------
Pierangelo Masarati wrote:
ando@OpenLDAP.org wrote:
provide an interface to remove a set of entries from the cache
The rationale behind this is that, for bandwidth occupation issues, we need to cache frequent proxied searches that retrieve very large objects. For this reason, we want relatively long TTL, but we also need to be very tight on updates on the remote server. For this purpose, I'm designing a layer based on the ldap_sync API (RFC 4533) that performs a persistent search looking for changes on the modifyTimestamp of the sensitive objects (small data that always gets modified, to be conservative) and, in case, removes those objects from the cache in a consistent manner. If this feature works as intended, and is considered useful, I can commit it (maybe it could become an integral part of pcache...).
Would the timestamp used need to be as high resolution as the one used in N-Way Multimaster?
Gavin.
Gavin Henry wrote:
Would the timestamp used need to be as high resolution as the one used in N-Way Multimaster?
No, in my design it's just an attribute that always gets modified. The value doesn't matter, since the purpose of this is to minimize the bandwitdh (but ldap_sync(3) does not know about delta-syncrepl) for the purpose of invalidating part of the cache.
p.
Ing. Pierangelo Masarati OpenLDAP Core Team
SysNet s.r.l. via Dossi, 8 - 27100 Pavia - ITALIA http://www.sys-net.it --------------------------------------- Office: +39 02 23998309 Mobile: +39 333 4963172 Email: pierangelo.masarati@sys-net.it ---------------------------------------
-
Gavin Henry -
Pierangelo Masarati