Hi
We have Notify 4.5.1 with 110 Blackberry users.
Since one month we have problem with the sync.
The sync works only from Scalix to Blackberry and not from Blackberry to Scalix.
Also a sync is very slow because to sync a user takes about 1 minute. 110 user x 1 minute=~2 hours.
Has somebody here in this forum experience with Scalix and Notify?
Thanks
Erich
Scalix with Notify
Moderator: ScalixSupport
-
jaime.pinto
- Scalix Star
- Posts: 709
- Joined: Fri Feb 23, 2007 6:50 pm
- Location: Toronto - Canada
Erich
You should probably call Nofity support and let them know about your scenario. There are many parameters that can be adjusted on the server and on the way the BB interacts with the server.
For example:
1) Keeping up with 52 weeks back and 52 weeks forward of of updates on calendars is a very demanding task even to 10 BB's. You could narrow that down to 2 weeks back and 8 weeks forward.
2) Some users set BB to check with the server every 2 minutes, which piles up on the server. You can overwrite that at system level to 5 minutes or even 10, which seems more appropriate in your case, if you have only 1 server for 100 users.
3) You can also adjust the amount of data and size of attachments to be sent back and forth.
4) There is also an overall priority direction, which seems to be your case. The directive is that the *NL server has to survive*. So the priority is to dispatch as much as possible from the server to the BB's and to the scalix server, at the cost of not taking incoming requests that could overload or choke the server. My impression is that you may need more than one server, which is fine, since the NL cost model is based on the number of BB/WinMob client licenses, not servers.
Just talk to them.
You should probably call Nofity support and let them know about your scenario. There are many parameters that can be adjusted on the server and on the way the BB interacts with the server.
For example:
1) Keeping up with 52 weeks back and 52 weeks forward of of updates on calendars is a very demanding task even to 10 BB's. You could narrow that down to 2 weeks back and 8 weeks forward.
2) Some users set BB to check with the server every 2 minutes, which piles up on the server. You can overwrite that at system level to 5 minutes or even 10, which seems more appropriate in your case, if you have only 1 server for 100 users.
3) You can also adjust the amount of data and size of attachments to be sent back and forth.
4) There is also an overall priority direction, which seems to be your case. The directive is that the *NL server has to survive*. So the priority is to dispatch as much as possible from the server to the BB's and to the scalix server, at the cost of not taking incoming requests that could overload or choke the server. My impression is that you may need more than one server, which is fine, since the NL cost model is based on the number of BB/WinMob client licenses, not servers.
Just talk to them.
Jaime||||||||||||||||||||||||||||||||||||||||
-
jaime.pinto
- Scalix Star
- Posts: 709
- Joined: Fri Feb 23, 2007 6:50 pm
- Location: Toronto - Canada
-
mnauta
- Posts: 152
- Joined: Wed Feb 23, 2005 4:10 pm
- Location: Austin, TX
Re: Scalix with Notify
aio wrote:Hi
We have Notify 4.5.1 with 110 Blackberry users.
Since one month we have problem with the sync.
The sync works only from Scalix to Blackberry and not from Blackberry to Scalix.
Also a sync is very slow because to sync a user takes about 1 minute. 110 user x 1 minute=~2 hours.
Has somebody here in this forum experience with Scalix and Notify?
Thanks
Erich
I have Notify Link On Demand. I had some major performance issue (5 users on Notify Link) about 80-100 scalix users with SWA at any given time. The %CPU was very high and IO waits very high. I installed new server this week, now %CPU for processes WAY down but IO waits still high.
What is your experience like? When the IO waits are high the users in both SWA and Outlook just sit there waiting for a response.
-
jaime.pinto
- Scalix Star
- Posts: 709
- Joined: Fri Feb 23, 2007 6:50 pm
- Location: Toronto - Canada
This thread probably is not the best place for this feedback, since the bad performance is mostly a function of the server (not a 3rd party integration issue), and not so much the clients of scalix, such at the Notifylink server.
We also went through the process of introducing a new server, more RAM, more/faster CPU, more HD space, newer OS. We definitely got an overall improvement, but at pick times we still saw the wait time high.
As it turns out, and it's well documented on several posts, Scalix is very I/O demanding on the mail storage.
If you're using "normal" hard drivers in a mirror scenario on the server you will see performance issue.
If your OS only allows for ext3 file system, such as RH4 or RH5, you'll see performance issue.
If you're using RAID5 on slow HDs you'll see performance issues.
The solution, and the IDEAL setup, is to use a very fast SAS raid and SAS HD's (not sata), and *definitely* not ext3 FS. You should use xfs, which will determine your choice of OS, to either ubuntu, centos with special kernel, or even xandros.
In our case we are now using iSCSI file system from a very fast raid (read/write of about 280-500MB/s) on the network, over gigE ports, and we practically get over 90% of the maximum bandwidth, ie, some 100-114MB/s. Incredibly enough, we still see some less than ideal performance problems early in the morning, when several users arrive at the office. Our next step up will be to replace the gigE card with 4gigE FC or 10gigE, possibly in April. Independently of that, ext3 is the worst offender on the equation, so just throwing hardware at the problem won't do the trick. We're hoping that in the next few months redhat will start to support xfs, otherwise we'll consult with scalix on a different option for OS.
I also think that scalix could consider some sort of *hybrid* mail store structure, ie, instead of one email one file on the file system, maybe we can have as many emails as it will fit on a 1MB file, similar to a small mbox or pst, and then move on to the next chunk of a file. Scalix performance is away more sensitive to the number of individual emails, in particular in the inbox, than to the size of the emails themselves. Apparently the number of calendar entries play a big role as well. As soon as a user logs in, the inbox and calendar are the first folders that get scanned and indexed. If a user keeps 2000 emails in the inbox and 1 full year of active calendar entries to be upgraded, there it goes your performance.
On the notifylink side, you need to determine how often you are setting up your users to push/pull information to/from the scalix server. If you have it set to 1 minute or 2 minutes then you're expecting too much. You'll be piling up update instructions on both the scalix and the Notifylink servers, and they won't be able to keep up. We have ours set to 5-8 minutes depending on the user, and this seems a good compromise.
We also went through the process of introducing a new server, more RAM, more/faster CPU, more HD space, newer OS. We definitely got an overall improvement, but at pick times we still saw the wait time high.
As it turns out, and it's well documented on several posts, Scalix is very I/O demanding on the mail storage.
If you're using "normal" hard drivers in a mirror scenario on the server you will see performance issue.
If your OS only allows for ext3 file system, such as RH4 or RH5, you'll see performance issue.
If you're using RAID5 on slow HDs you'll see performance issues.
The solution, and the IDEAL setup, is to use a very fast SAS raid and SAS HD's (not sata), and *definitely* not ext3 FS. You should use xfs, which will determine your choice of OS, to either ubuntu, centos with special kernel, or even xandros.
In our case we are now using iSCSI file system from a very fast raid (read/write of about 280-500MB/s) on the network, over gigE ports, and we practically get over 90% of the maximum bandwidth, ie, some 100-114MB/s. Incredibly enough, we still see some less than ideal performance problems early in the morning, when several users arrive at the office. Our next step up will be to replace the gigE card with 4gigE FC or 10gigE, possibly in April. Independently of that, ext3 is the worst offender on the equation, so just throwing hardware at the problem won't do the trick. We're hoping that in the next few months redhat will start to support xfs, otherwise we'll consult with scalix on a different option for OS.
I also think that scalix could consider some sort of *hybrid* mail store structure, ie, instead of one email one file on the file system, maybe we can have as many emails as it will fit on a 1MB file, similar to a small mbox or pst, and then move on to the next chunk of a file. Scalix performance is away more sensitive to the number of individual emails, in particular in the inbox, than to the size of the emails themselves. Apparently the number of calendar entries play a big role as well. As soon as a user logs in, the inbox and calendar are the first folders that get scanned and indexed. If a user keeps 2000 emails in the inbox and 1 full year of active calendar entries to be upgraded, there it goes your performance.
On the notifylink side, you need to determine how often you are setting up your users to push/pull information to/from the scalix server. If you have it set to 1 minute or 2 minutes then you're expecting too much. You'll be piling up update instructions on both the scalix and the Notifylink servers, and they won't be able to keep up. We have ours set to 5-8 minutes depending on the user, and this seems a good compromise.
Jaime||||||||||||||||||||||||||||||||||||||||
-
PrisonMind
- Posts: 308
- Joined: Thu Jul 10, 2008 5:00 am
- Location: Adua
we installed a centos plus kernel + headers and xfs tools, so you can use rhel5 with xfs.
but only for test...we didnt use this setup on a production server!
so its your own risk!
or you wait..
ext4 its a part of rhel6 ( is going to be Apr / May 2009 )
@jamie
have you got experience with redhats gfs/gfs2??
but only for test...we didnt use this setup on a production server!
so its your own risk!
or you wait..
ext4 its a part of rhel6 ( is going to be Apr / May 2009 )
@jamie
have you got experience with redhats gfs/gfs2??
-
ECfoS
Hello
On Scalix is normal to see a performance problem
Scalix use a "logon-logout" on each user
an upgrade of Scalix to support caldav sync of Calendar and
once this is done you can make some fine tuning also on NL server
after that , you will see a beter performance on the sync
to know more about this...ping me directly
best regrads
Sergio
On Scalix is normal to see a performance problem
Scalix use a "logon-logout" on each user
an upgrade of Scalix to support caldav sync of Calendar and
once this is done you can make some fine tuning also on NL server
after that , you will see a beter performance on the sync
to know more about this...ping me directly
best regrads
Sergio
-
billb3
- Scalix Star
- Posts: 464
- Joined: Mon May 26, 2008 8:56 pm
- Location: Kingston, NY
- Contact:
Re:
jaime.pinto wrote:We're hoping that in the next few months redhat will start to support xfs, otherwise we'll consult with scalix on a different option for OS.
It appears that RHEL/CentOS 5.4 will support XFS. I for one am planning to switch from ext3
http://dag.wieers.com/blog/red-hat-back ... g-to-rhel5
Return to “Third Party Integration”
Who is online
Users browsing this forum: No registered users and 2 guests