Hi Linda/Dan,
We need to know if there is something that can be done to improve our the performance for our internet/vpn users. Here are our details:
Server: AMD 3500+, 2Gb RAM, SQL Server 2000 SP3a, Windows 2003 Standard (no SP). Vault server 3.1.2.
Database: ~20GB, possibly 30-40% can be obliterated but slow process.
Clients: Mainly WinXP pro (SP2), Vault client 3.1.2.
Internal users do not have any issues.
External internet or VPN users exhibit slow performance to retrieve files. Usually a 1MB file can take 10 mins to download based on a 512KB line. Internally this is a few seconds.
Main repository as you know is pretty big, 17527 folders (11638 deleted) / 150547 files (109295 deleted). Disk space needed 53GB, Database size 13GB. Treesize, 26GB.
Smaller repositories have no issues, so we think this rules out the clients bandwidth. We think the server side is causing the issues. Do you have any suggestions?
Performance problems with 3.1.2.
Moderator: SourceGear
we have a similiar problem, details follow; is there any solution?
Server: dual xeon 2.8ghz, 2gb ram, sql 2000 sp3, windows server 2000, vault 3.5.1.4786
database 9857mb, 4686mb available
34794 files, 2578 folders in this repository
Internal users do not have any issues.
a GLV that only takes me a minute or two to process (creating a project folder of ~ 65mb) might take one of our remote users 75 minutes.
the database buffer size is set to 256 but the help file says default is 768, could this have an effect?
reverse DNS is off
vault apparently sends data to IIS for delivery to the client in 10MB chunks, and this seems helpful in diagnosing slowness, because the vault log has this interesting content (simplified for posting):
remote user:
7:05:29 AM Getting repository Structure-> Rep ID: 1 Base: 35681 Target: -1
7:05:29 AM GetRepositoryStructure returned: Success
7:05:32 AM Getting list of checkout changes.
7:05:32 AM GetCheckOutListChanges returned: Success
7:05:39 AM Beginning file download
7:05:39 AM BeginDownloadFiles returned: Success
7:06:35 AM VaultFileDownload starting
7:06:36 AM GetLatest wrote 10756818 bytes to the Response Stream
7:25:19 AM GetLatest wrote 11148219 bytes to the Response Stream
7:54:12 AM GetLatest wrote 10038355 bytes to the Response Stream
8:32:09 AM Ending download process
8:32:09 AM EndDownloadProcess returned: Success
local user:
8:29:48 AM Getting repository Structure-> Rep ID: 1 Base: 35719 Target: -1
8:29:48 AM GetRepositoryStructure returned: Success
8:29:48 AM Getting list of checkout changes.
8:29:48 AM Beginning file download
8:29:48 AM BeginDownloadFiles returned: Success
8:29:49 AM VaultFileDownload starting
8:29:50 AM GetLatest wrote 10756818 bytes to the Response Stream
8:30:28 AM GetLatest wrote 10043353 bytes to the Response Stream
8:31:09 AM Ending download process
8:31:09 AM EndDownloadProcess returned: Success
notice how there is mere seconds between each 10mb write for the local user, but 20-30 minutes for each 10mb write for the remote users. the remote users can download (via ftp) from the same vault server at approximately 115kB/sec, or approximately 6mb a minute, so 10mb taking 20-30 minutes is really excessively long.
any ideas???
Server: dual xeon 2.8ghz, 2gb ram, sql 2000 sp3, windows server 2000, vault 3.5.1.4786
database 9857mb, 4686mb available
34794 files, 2578 folders in this repository
Internal users do not have any issues.
a GLV that only takes me a minute or two to process (creating a project folder of ~ 65mb) might take one of our remote users 75 minutes.
the database buffer size is set to 256 but the help file says default is 768, could this have an effect?
reverse DNS is off
vault apparently sends data to IIS for delivery to the client in 10MB chunks, and this seems helpful in diagnosing slowness, because the vault log has this interesting content (simplified for posting):
remote user:
7:05:29 AM Getting repository Structure-> Rep ID: 1 Base: 35681 Target: -1
7:05:29 AM GetRepositoryStructure returned: Success
7:05:32 AM Getting list of checkout changes.
7:05:32 AM GetCheckOutListChanges returned: Success
7:05:39 AM Beginning file download
7:05:39 AM BeginDownloadFiles returned: Success
7:06:35 AM VaultFileDownload starting
7:06:36 AM GetLatest wrote 10756818 bytes to the Response Stream
7:25:19 AM GetLatest wrote 11148219 bytes to the Response Stream
7:54:12 AM GetLatest wrote 10038355 bytes to the Response Stream
8:32:09 AM Ending download process
8:32:09 AM EndDownloadProcess returned: Success
local user:
8:29:48 AM Getting repository Structure-> Rep ID: 1 Base: 35719 Target: -1
8:29:48 AM GetRepositoryStructure returned: Success
8:29:48 AM Getting list of checkout changes.
8:29:48 AM Beginning file download
8:29:48 AM BeginDownloadFiles returned: Success
8:29:49 AM VaultFileDownload starting
8:29:50 AM GetLatest wrote 10756818 bytes to the Response Stream
8:30:28 AM GetLatest wrote 10043353 bytes to the Response Stream
8:31:09 AM Ending download process
8:31:09 AM EndDownloadProcess returned: Success
notice how there is mere seconds between each 10mb write for the local user, but 20-30 minutes for each 10mb write for the remote users. the remote users can download (via ftp) from the same vault server at approximately 115kB/sec, or approximately 6mb a minute, so 10mb taking 20-30 minutes is really excessively long.
any ideas???