Professional Documents
Culture Documents
Technical Notes
P/N 300-011-878 REV A01 November 16, 2010
Executive summary ................................................................................... 2 Introduction ................................................................................................ 2 Virtual device restore: Additional SAVE area space considerations .. 2 Adding restore considerations into sizing of the SAVE area ............... 5 Conclusion .................................................................................................. 7
Executive summary
Executive summary
The EMC TimeFinder SNAP Facility is a space-saving local replication technology. Part of implementation planning requires calculating the change rate of the source devices to be SNAPPED in order to calculate the size of the SAVE area. This is typically achieved using the change tracker tool. This document describes how extra data may be written to the SAVE area when a SNAP restore is performed and details the SAVE area planning considerations for SNAP restore.
Introduction
This technical notes document discusses the additional SAVE area overhead that can be generated by performing SNAP restore operations in a multi-SNAP environment; it further illustrates how to calculate how much overhead (number of tracks) will be created before a restore is performed. As there can be extra SAVE space used in restore processing, the final section in this document describes how to include restore processing overhead when calculating SAVE area requirements.
Audience
This document is intended for users and planners of TimeFinder SNAP technology.
SAVE Area Space Considerations for Virtual Device Restore Technical Notes
To complete the restore, SNAP sessions must be terminated to free up space in the SAVE area or more SAVE space needs to be added. Do not terminate the SNAP being used for the restore! Once the SAVE area has been made available, the restore automatically continues until successful completion.
C:\Temp>symsnap -g testSC query multi Device Group (DG) Name: testSC DG's Type : REGULAR DG's Symmetrix ID : 000287970245 Source Device Target Device State Copy ------------------------- ------------------- ---------- ------------ --Protected Changed Logical Sym Tracks Logical Sym G Tracks source <=> TGT (%) ------------------------- ------------------- ---------- ------------ --DEV001 009C 137218 136390 135559 134727 -------543894 16996.7 VDEV004 VDEV003 VDEV002 VDEV001 00A1 00A0 009F 009E X X X X 842 1670 2501 3333 ---------8346 260.8 CopyOnWrite CopyOnWrite CopyOnWrite CopyOnWrite 0 1 1 2
SAVE Area Space Considerations for Virtual Device Restore Technical Notes
2.
C:\Temp>symsnap -sid 245 -savedevs list Symmetrix ID: 000287970245 S N A P S A V E D E V I C E S --------------------------------------------------------------------Device SaveDevice Total Used Free Full Sym Emulation Pool Name Tracks Tracks Tracks (%) --------------------------------------------------------------------009A FBA Pool_Test 138060 3371 134689 2 Total --------- --------- -----------Tracks 138060 3371 134689 2 MB(s) 4314.4 105.3 4209.0
3.
Determine which VDEV is to be restored from. In this example we will restore from the oldest SNAP, which is VDEV001. Note the changed tracks (3333). The newest SNAP (VDEV004) used 842 tracks. Calculate the difference between the VDEV restore source and the newest SNAP: 3333 842 = 2491 tracks to be added to the SAVE area. The tracks used in the save area in the output above is 3371. After the restore the SAVE area should contain 3371 + 2491 = 5862 tracks.
4.
SAVE Area Space Considerations for Virtual Device Restore Technical Notes
--------------------------------------------------------------------009A FBA Pool_Test 138060 5862 132198 4 Total --------- --------- -----------Tracks 138060 5862 132198 4 MB(s) 4314.4 183.2 4131.2
The SAVE area used before restore was 3371 tracks, After restore it was 5862 for a difference of 2491 tracks. This matches exactly the pre-restore calculation.
What do I do if the restore space calculation exceeds the available free space?
Preserving some free space in the SAVE areas is of paramount importance as filling the SAVE area FAILS ALL active SNAPs once they are written to when the SAVE area is full. If the restore space calculation exceeds the available free space, either more space is needed in the SAVE pool or some SNAP sessions must be terminated to increase available free space in the SAVE pool.
The additional SAVE area restore overhead may be 0 percent to 100 percent depending on the use of SNAP. In order to include restore space into the SAVE area calculations, perform the following: Determine the SAVE area requirements for the SNAPS as is currently best practice Of the total number of VDEVS to be used, determine the percentage that will be used as multi-SNAPs (more than one VDEV active to a standard device concurrently) Of the number of source devices using multi-SNAPs, estimate the percentage that may need to be restored concurrently Discount the amount of SAVE area used by the latest SNAP as this is
5
SAVE Area Space Considerations for Virtual Device Restore Technical Notes
not written into the SAVE area in a restore. Calculate the restore area using the details above as follows: TOTAL save area = Initial SAVE area calculation + (Initial SAVE area calculation x (# multi-SNAP VDEVs /total VDEVs ) x (# concurrent restore source devices x total multi-SNAP source devices)) Discount the space from the latest SNAP. The amount of space used by the latest SNAP as a percentage of the total space used for all VDEVS in a multi-SNAP is dependent on the number of multi-SNAPs being used by the source device(s). The following table is a guide to the percentage of space used by the latest SNAP in a multiSNAP environment. Table 1 Percentage of space used by the latest SNAP based on number of SNAPs Number of SNAPs per source device 2 3 4 5 6 7 8 9 10 Percentage of space used by the latest SNAP 25 11 6 4 3 2 2 1 1
Conclusion
In this example, including the restore space has almost doubled the initial SAVE area requirement. Example No. 2 The SAVE area for a new SNAP implementation was calculated to be 100 GB. There are a total of 400 VDEVs and 100 source devices. Twenty of the source devices are to have five SNAPs active at any point. The remainder will only have one active SNAP. If there is a requirement to restore from a VDEV to the source device, half (10) of the devices with multi-SNAP sessions will need to be restored concurrently. 100 GB + (100 GB x ( 80/400) x (10/20)) = 110 GB 110 GB 4 percent = 105.6 GB In this example, including the restore space has added nearly 10 percent to the SAVE area requirement.
Conclusion
Performing SNAP restores in an environment that uses multi-SNAPs does write additional data into the SAVE area. The amount of SAVE overhead depends on the implementation and can be as little as 0 percent or as much as 100 percent. When sizing the SAVE area it is recommended to consider the additional sizing criteria described in this document in order to ensure a quality SNAP implementation.
SAVE Area Space Considerations for Virtual Device Restore Technical Notes
Conclusion
SAVE Area Space Considerations for Virtual Device Restore Technical Notes