DAOS rocks!
Category Lotus Notes and Domino 8.5 DAOS
Bookmark :
Just recently we implemented DAOS on our three clustered mail servers, and our lone mail archive server. The implementation went without any problems, and the outcome is really amazing. Here's why...
As a real life example of the difference that DAOS makes, let's examine the mail file size of one of our typical users. This user, just like others here, has his mail file replica on each of the three clustered servers. Yesterday, there were 18,740 documents (many with attachments) in each replica. While the replicas on servers ZEUS and VENUS were already converted to ODS 51 (required for DAOS), I noticed that the replica on server TRITON was still using ODS 43, hence DAOS did not do its magic here (see fig.1).
Mail file size on ZEUS was 1,354 MB with 97.5 of space used (ie. 2.5 % white space).
Mail file size on VENUS was 1,345 MB with 97.8 of space used (ie. 2.2 % white space).
Mail file size on TRITON was 3,156 MB with 99.2 of space used (ie. 0.8 % white space).
Do you see the difference? 1.3 GB on a server with DAOS versus 3.1 GB on a server without DAOS. Just by converting the mail file replica on TRITON to ODS 51 and enabling DAOS on it (compact -c), I brought the size down to 1.3 GB (see fig2). I just recovered 1.8 GB of disk space from one mail file.
Fig 1. DAOS at work on ZEUS and VENUS mail file replica (ODS 51), but not TRITON (ODS 43)
Fig. 2 DAOS at work on all three replicas of the mail file after the replica
on TRITON has been converted to ODS 51
How was this possible? This user had a lot of attachments stored in his mail file, that were also stored in other user's mail files. When DAOS went thought that mail file, it stripped duplicated attachments from the documents and replaced them with pointers to those same files that DAOS already stripped form other mail files and stored in one central location on the server.
Overall, we recovered about 200 GB of storage space on each clustered server after implementing DAOS. 3 x 200 GB, that's 600 GB of space back to the pool.
Now, the mail archive server. We have about 2,700 users with mail files. Each user has a replica on each of the three mail servers in the cluster. Any mail file that exceeds 400 MB, is eventually included in scheduled server-side archiving to a dedicated mail archive server called HADES. As you can imagine, HADES needs a lot of storage space to accommodate all these archives, especially because we do not have mail file quotas in place and users are allowed to keep anything they want in their mailboxes. Before we implemented DAOS on HADES, the server was using 1.7 TB of space on our SAN for mail file archives. After we implemented DAOS on all mail archives on HADES, we recovered 600 GB (thanks Patrick...) of storage space on that partition. That's more that 1/3 of the space previously used to store these files.
And here is the final proof that DAOS is just amazing. Take good look at the pictures below. This is Chris, our SAN and backup administrator. The first picture was taken just before DAOS was implemented. The second picture was taken after DAOS was running. What a difference!
Fig. 3 Chris before DAOS.
Fig. 3 Chris after DAOS.
Bookmark :
Just recently we implemented DAOS on our three clustered mail servers, and our lone mail archive server. The implementation went without any problems, and the outcome is really amazing. Here's why...
As a real life example of the difference that DAOS makes, let's examine the mail file size of one of our typical users. This user, just like others here, has his mail file replica on each of the three clustered servers. Yesterday, there were 18,740 documents (many with attachments) in each replica. While the replicas on servers ZEUS and VENUS were already converted to ODS 51 (required for DAOS), I noticed that the replica on server TRITON was still using ODS 43, hence DAOS did not do its magic here (see fig.1).
Mail file size on ZEUS was 1,354 MB with 97.5 of space used (ie. 2.5 % white space).
Mail file size on VENUS was 1,345 MB with 97.8 of space used (ie. 2.2 % white space).
Mail file size on TRITON was 3,156 MB with 99.2 of space used (ie. 0.8 % white space).
Do you see the difference? 1.3 GB on a server with DAOS versus 3.1 GB on a server without DAOS. Just by converting the mail file replica on TRITON to ODS 51 and enabling DAOS on it (compact -c), I brought the size down to 1.3 GB (see fig2). I just recovered 1.8 GB of disk space from one mail file.
Fig 1. DAOS at work on ZEUS and VENUS mail file replica (ODS 51), but not TRITON (ODS 43)
Fig. 2 DAOS at work on all three replicas of the mail file after the replica
on TRITON has been converted to ODS 51
How was this possible? This user had a lot of attachments stored in his mail file, that were also stored in other user's mail files. When DAOS went thought that mail file, it stripped duplicated attachments from the documents and replaced them with pointers to those same files that DAOS already stripped form other mail files and stored in one central location on the server.
Overall, we recovered about 200 GB of storage space on each clustered server after implementing DAOS. 3 x 200 GB, that's 600 GB of space back to the pool.
Now, the mail archive server. We have about 2,700 users with mail files. Each user has a replica on each of the three mail servers in the cluster. Any mail file that exceeds 400 MB, is eventually included in scheduled server-side archiving to a dedicated mail archive server called HADES. As you can imagine, HADES needs a lot of storage space to accommodate all these archives, especially because we do not have mail file quotas in place and users are allowed to keep anything they want in their mailboxes. Before we implemented DAOS on HADES, the server was using 1.7 TB of space on our SAN for mail file archives. After we implemented DAOS on all mail archives on HADES, we recovered 600 GB (thanks Patrick...) of storage space on that partition. That's more that 1/3 of the space previously used to store these files.
And here is the final proof that DAOS is just amazing. Take good look at the pictures below. This is Chris, our SAN and backup administrator. The first picture was taken just before DAOS was implemented. The second picture was taken after DAOS was running. What a difference!
Fig. 3 Chris before DAOS.
Fig. 3 Chris after DAOS.
Comments
Posted by ric At 08:31:39 AM On 12/30/2009 | - Website - |
Posted by tina kandel At 07:37:22 AM On 11/05/2009 | - Website - |
Posted by Claude Cossette At 11:27:04 AM On 09/25/2009 | - Website - |
Posted by Paul Mooney At 05:17:21 AM On 06/20/2009 | - Website - |
Posted by Bodek Frak At 11:13:46 AM On 06/19/2009 | - Website - |
Before we implemented DAOS on HADES, the server was using 1.7 TB of space on our SAN for mail file archives. After we implemented DAOS on all mail archives on HADES, we recovered <<<<600 MB>>>>> of storage space on that partition. That's more that 1/3 of the space previously used to store these files.
Posted by Patrick Picard At 09:39:25 AM On 06/19/2009 | - Website - |