or the theme Staging Folder. Staging quota is an often neglected topic, resulting in consequence, however in practice the most problems. It is in this howto realize a connection line between these two independent settings drawn, as between the size of the staging folder (So the Quota) and the replication behavior, there is a strong connection.
First, a little theory, the very roughly sketched, what happens when you change.
For example, when a file on Server A changed, replicates its data to the server B, sends server A after the change a "change notification" to the server B. After comparing the databases (or. whose entries) is Server B now start an inquiry to Server A, the file to be replicated.
Server A is now, after receiving the request from server B, the corresponding file "trainee", d.h. grab the file and put it into the staging folder below "DfsPrivateStaging".
There is using Remote Differential Compression (DRC) checked, what parts need to be transferred to the file, and the file is then broken down into the necessary parts.
These parts are transferred to the server B, composed the file again (possibly. involving the local parts of the file, that were not transferred with, because they were already present on the server B), to then extracted to create the location.
This process costs, depending on file size and file type, a lot of CPU power and disk performance. It therefore makes sense, to gestagen re not always just with often changed files data, but to leave the files are possible in the Staging Folder, since there are already calculated many "file pieces".
The staging folder is subject to the staging quota. That means, that the server A and the server which have B configures a staging quota determined, how much data can be put there.
If this quota is too low, must be gestaged anew the files repeatedly, which can have a high CPU load and disk load result. Moreover, the data transfer rate decreases and may, at worst, to almost a complete stop replication.
This is caused Settings, uses the Microsoft for the shrouds of data by default. There is a so-called low-water mark (60% Staging of-Quota) and a High Water Mark (90% Staging of-Quota). If the high water mark is exceeded, to delete the oldest data from the Staging Folder, to the low-water mark is reached again, also 60% quotas.
Exceeding the watermarks and the resulting deletion of data is in the DFS-R Eventlogs among the Event IDs 4202, 4204, 4206 and 4208 appropriate. A sensible measure from this point could therefore be, the quota by at least 50% to increase.
So If you imagine, that at a staging quota of 4096 MB a ca. 4 GB would gestagen large file…
– it must packed,
– calculates checksums
– be dismantled and;
– it must be deleted from the Staging Folder almost all data,
– then the 4 transfer GB file.
After this mechanism to other files will also be held again, which in turn must be recalculated - a high expenditure.
As long as the staging folder size under 100% of quotas is, is the DFS-R Server replicate nine files at a time: 5 sending (outbound) and 4 received (inbound) Threads, d.h. a total of 9 files.
If a quota is reached (the quota is "elastic", not "static"), is a thread for deletion used, until the quota utilization of affected replication group again under 60% sinks.
However, at the time of the quota are exceeding all 4 inbound or all 5 used outbound threads, may occur in unfavorable constellations to, that ALL outbound or. inbound threads or utilized RPC connections the server be blocked, complete scavenging of Quotas. This means in practice, that hardly or. any other data other than this one file and increase the so-called strong backlogs.
To illustrate again: Occurs this effect, can on the entire server no replication group replicate more data, until the quota again under 60% declined. However, if a file in the Staging Folder, which is a whole greater than the quota, can last until the completion of replication of this file, this process. This means in practice therefore a temporary stoppage of data replication from this server. Other models to be found at [6].
The quotas have by default in new replication folders size of 4096 MB. This value should be adapted in any case his claims. It must be clearly stated, that there can be no concrete statement to the size of quotas in principle – this is to decide in individual cases. However, one should the size of the quotas if possible as large as humanly possible in proportion to the mass of data the data to be replicated choose. At best, the quota is even the same as the data of the corresponding mass replication group.
During the initial replication much quota space is required in the normal case, since all data must be transferred (unless it were data Pre-Staged or cross-file RDC is possible). For this reason, it is recommended, to choose very generous during the initial replication the Quota. In the best case, if enough memory, at least as large as the total amount of data to be replicated (see above).
If you have more free memory, you can even up to a thumb formula of 1.5x replication folder size go for the Quota, because by checksums, old staging files etc. certainly can come together more data, provides as the folder itself.
It should be noted, that the quotas (briefly mentioned above) or on all servers. Replication folders must be set individually. Furthermore is noted, that the quotas apply only per replication folder, ie on all replication groups as needed can be customized (and should).
NOTE: At the end of this thread once again the haunting Note, very good to plan quotas and always keep in mind. Are there problems here, is sometimes all DFS-R structure affected.
Author: bad, MCSEboard.de