Académique Documents
Professionnel Documents
Culture Documents
Table of Contents
DXi-Series Best Practices for NetBackup ................................................................... 1
Scope ....................................................................................................................................... 1
Quantum Corporation
Page i
Scope
The intended audience for this guide is our customers and field sales teams as guidance to enhance the installation, integration, and utilization of the Quantum DXi-Series. Specifically, this guide provides recommendations and considerations impacting installing, integrating, and utilizing the DXi-Series products specifically for NetBackup by Symantec environments.
Quantum Corporation
Page 1
Installation Recommendations
Virtual Tape Library Presentation: Fibre Channel
Robot/Media Changer Device Serialization and Recommendations
Quantum is a leading manufacturer of automation devices and extends this knowledge and experience to the Virtual Tape Library (VTL) disk-based solutions. One of the key components to ensure that SANconnected physical and virtual tape libraries are detected properly by backup servers is serialization. Serialization provides a unique identifier for each device in a physical or virtual tape library to automate device association from multiple backup servers. These identifiers, returned by the VTL devices, are separate from the element address that defines the position of devices in the library. The element address is used by the librarys robot or medium changer to manage the tape drives. Serialization allows the servers running the data protection application (the media servers) to coordinate tape drive configuration by aligning the device serial number with the devices element address. This enables the data protection application to automatically discover the devices in the library during the device discovery and configuration phase (e.g., the NetBackup Device Configuration Wizard), reducing the potential for improper configuration. This feature enables Symantec NetBackup device discovery to align these two addresses. If the Device Configuration Manager does not serialize the devices listed, do not commit the changes, and check the VTL online state. The DXi VTL partition must be online for this to function properly. The Quantum recommended device identification for each DXi system is the native mode (e.g., use the DXi6700, DXi7500 or DXi8500 inquiry response string as the identification for each model respectively). This allows product identification for the service teams at both Symantec and Quantum. When using the native device mode, Windows environments will display the device in the device manager as an unknown media changer. This is normal and not an error or problem to NetBackup. If the customer environment has requirements for a specific changer device for compatibility, the Quantum DXi products support emulation of many popular devices to meet those requirements. Please note NetBackup requires that VTLs be configured with the native inquiry string. Emulation stings are not supported. Quantum recommends installing NetBackups latest device mapping file to support properly identifying devices in NetBackup.
Quantum Corporation
Page 2
Virtual Tape Library operation can be tested with the robtest utility included with NetBackup. Details of the robtest utility can be downloaded form the NetBackup support site: http://seer.entsupport.symantec.com/docs/274063.htm
Quantum Corporation
Page 3
Notice that a single FC Node could reside in multiple zones. This is due to the fact that it has multiple targets to contact.
Quantum Corporation
Page 4
among multiple media servers to simplify initial installation by providing dedicated resources to each media server. Quantum recommends that backups be staggered such that only a subset of drives are in use at one time. During a backup the data transfer rate is primarily controlled by the media server because the DXi system does not restrict the ingest data rate. This creates the opportunity for one or more media servers to burst data at a higher rate, leaving the remaining virtual tape drives less bandwidth. Conversely, it supports the coexistence of fast data streams with slow streams for maximum utilization of the available bandwidth. Note: Increasing the number of concurrently active virtual tape drives does not increase the aggregate DXi bandwidth and could result in a failed backup job due to a timeout from a bandwidth-starved operation. The table below lists the recommended maximum number of concurrently active virtual tape drives for the respective system bandwidth.
DXi Model DXi6700 DXi6701/DXi6702 DXi7500 Express DXi7500 Enterprise DXi8500 Maximum Virtual Tape Drives 80 80 80 160 160 Maximum Aggregate Bandwidth 3.5TB/Hr or 972 MB/sec 5.8TB/Hr or 1610 MB/sec 1.28TB/Hr or 355 MB/sec 4TB/Hr or 1,111MB/sec 6.4TB/Hr or 1,777MB/sec Recommended Maximum Number of Concurrently-Active Virtual Tape Drives Up to 48 concurrently-active virtual tape drives ** Up to 80 concurrently-active virtual tape drives ** Up to 17 concurrently-active virtual tape drives ** Up to 55 concurrently-active virtual tape drives ** Up to 88 concurrently-active virtual tape drives**
** Based on a user wanting no less than20 MB/sec per virtual tape drive.
Quantum Corporation
Page 5
Be aware that capacity utilization for DXi8500, DXi7500, DXi6700 and DXi6701/DXi6702 is tracked in COMPRESSED GB and the data is stored compressed. That is, 100GB of data that is 2:1 compressible will report that it occupies 50GB of virtual tape cartridge space.
If Application Specific Path to Tape is utilized, the ultimate cartridge destination size is the suggested size. Quantums general guidance is to specify a smaller virtual tape cartridge capacity (such as 100GB) for the reasons mentioned above.
Quantum Corporation
Page 6
Note: The NAS shares created on the DXi system are not recommended for general-purpose storage. Quantum DXi systems support both CiFS (Windows-based) and NFS shares. Each system can support multiple NAS shares with a maximum of 128 shares. Quantum recommends that users create only the required number of shares for each media server. DXi systems can support concurrent NFS and CiFS shares and can support Fibre Channel VTLs concurrently with those NFS and CiFS shares. When using NAS shares on the DXi systems, Quantum recommends creating at least one share for each media server to use. Media servers should not share the NAS shares during normal backup operations. Root access to an NFS share is not allowed, and the access rights will be changed to nfsnobody as a security precaution. This does not impact access to the share from the NetBackup application.
Replication Recommendations
For first-time replication setup, it is important to manually replicate the name space once the target system is configured and is online. This facilitates the first replication following the first backup to that share/partition. The replication is only available to VTL and NAS shares with de-duplication enabled. The DXi supports 128-bit AES encryption for replication. Data is only encrypted while in transit between replication source and replication target. Data is unencrypted upon arrival at the replication target. Encryption may affect replication performance. It should be disabled if the customers WAN is already secured. Please refer to the Quantum DXi-Series Best Practices for Data Replication (Document 20020430-xxx).
Quantum Corporation
Page 7
Integration Recommendations
NetBackup with DXi Series Configuration Guide
The NetBackup with DXi Series Configuration Guide contains additions additional information about configuring solutions with NetBackup (see the DXi-Series Configuration Guide with NetBackup PN 20020429-xxx)
Quantum Corporation
Page 8
Performance Gains with Buffer Settings in NetBackup Three main settings in NetBackup have been shown to have a major performance impact for backups: the size of the data buffer, the number of the data buffers, and the size of the network data buffers. SIZE_DATA_BUFFERS After changing this value, it is important to test both backups and restores, because sometimes data can be written at the modified size but potentially cannot be read at the modified size. Our lab environment testing and field experience has been positive using the recommended 262,144 block size. To change the SIZE_DATA_BUFFERS setting: 1. Create a file called SIZE_DATA_BUFFERS in the <INSTALL_PATH>\\NetBackup\\db\config. Note: The file name is case sensitive and must have no extension. 2. Add the value 262144 in this file. Note: If the file is not present, the default value for the SIZE_DATA_BUFFER is 65,536. Windows-specific guidance to use 256K block size: Windows Server 2003 requires that SP2 be installed; otherwise, the maximum allowed is 64K. After this file is created you do not have to cycle any services. The change will be done when the next backup job starts. In combination with the change in block size, the number of buffers used will impact overall system performance. The default setting for the number of buffers used is 16. Increasing the number of buffers al-
Quantum Corporation
Page 9
lows more data to be buffered before being sent to the device. Our lab testing has yielded better results using 32 for the number of buffers. NUMBER_DATA_BUFFERS To change the NUMBER_DATA_BUFFERS setting: 1. Create the <INSTALL_PATH\NetBackup\db\config\NUMBER_DATA_BUFFERS file. Note: The file name is case sensitive and must have no extension. 2. Add the value 32 in this file Note: If the file is not present, the default value 16 will be used. After this file is created you do not have to cycle any services. The change will be done when the next backup job starts. Network Buffering For Windows Servers Overview When a backup is initiated, the client packages data of the amount specified by the Buffer_size value, and then transfers the information to the media server, which in turn buffers that data in the NET_BUFFER_SZ. When the NET_BUFFER_SZ is full, it transfers data to the array of space created by a combination of NUMBER_DATA_BUFFERS and SIZE_DATA_BUFFERS. As soon as at least one of the SIZE_DATA_BUFFERS is full, the information is written to the tape drive. NET_BUFFER_SZ Note the NET_BUFFER_SZ value is the size of the buffer on the media server that receives data from the client server. To change the Net_Buffer_SZ: 1. Create the NET_BUFFER_SZ file in the <INSTALL_PATH>\NetBackup directory. Note: The file name is case sensitive and must have no extension. The value in the file must be a multiple of 1024. If the file is not present, the default value 262144 (or 256K) is used. 2. Add the appropriate value to this file. Our lab testing has yielded improved performance by matching the Buffer_Size on the clients to NET_BUFFER_SZ on the media server. Buffer_size Note: The value of the Buffer_size does not show up in the bptm log file, but can be viewed and set from the master server or by modifying a registry value. The default value for the Buffer_size setting is 32K. Changing the client server, the buffer size (Buffer_size) can be modified from the application user interface. Changing the Client Buffer_size (called communications buffer):
Quantum Corporation
Page 10
From the NetBackup Administration console, Host Properties | Clients, double-click the client you are configuring. Expand the options under Windows Client | select Client Settings. Change the Communications buffer to the desired size in KB.
For additional information on the settings, see our partner site link: http://support.veritas.com/docs/244652
Quantum Corporation
Page 11
Quantum Corporation
Page 12
Use the largest rsize/wsize that both the client and server support since the client and server will negotiate down. Use 1048576 (the default for many clients) since this is the maximum supported by DXi. 1. Mount with rsize/wsize set to 1M.
mount o vers=3,tcp,rsize=1048576,wsize=1048576
2. Verify the mount options by running the command cat /proc/mounts NOTE: The older versions of Linux (pre-2.6.16) only support maximum of 32K rsize/wsizes that are not optimal. 2.6.18 and recent versions showed better performance.
Step # 1
3.
Quantum Corporation
Page 13
OPTIONAL. If your client has multiple NICs bonding can be beneficial. Create Link aggregation group with LACP.
Solaris# dladm create-aggr d bge0 d bge1 1 Solaris# ifconfig aggr1 unplumb Solaris# dladm modify-aggr P L4 l active 1 Solaris# ifconfig aggr1 plumb <ip> netmask <netmask> up
Step# 1
2.
Maximize parallelism between an NFS client and its servers; set the following in the /etc/system file
set rpcmod:clnt_max_conns = 8
3.
4. 5.
Maximize the contiguous blocks on the NFS share with the following command
tunefs a 256 /<SHARENAME>
Quantum Corporation
Page 14
For NAS configurations set socket buffer settings to 4M default. Append these lines to /etc/sysctl.conf
net.ipv4.tcp_rmem = 4194304 4194304 4194304 net.ipv4.tcp_wmem = 4194304 4194304 4194304 net.ipv4.tcp_mem = 10485760 10485760 10485760 net.core.netdev_max_backlog = 30000 Run the command sysctl p for the newly added configuration to take effect. Verify the configuration by running sysctl net.ipv4.tcp_rmem, it should show 4194304 4194304 4194304. Note: Set the socket buffers before doing mounts from clients. If they are already mounted, clients need to unmount and remount for the settings to take place. The unmount and remount may be done after the client is fully configured.
OPTIONAL. If your client has multiple NICs bonding can be beneficial. Create Link aggregation group with LACP
And add bonded round robin policy by adding this line in /etc/modprobe.conf: options bonding mode=6 miimon=500 [NOTE: During this testing, in lieu of LACP, Layer3+4 xmit_hash policy was also successfully used to reduce TCP retransmissions. In the /etc/modprobe.conf, add the line options bonding mode=2 xmit_hash_policy=layer3+4 miimon=500 NOTE: this step needs a reboot.
Step# 1
Linux NFS Client Settings Mount with rsize/wsize set to 1M. This testing almost always had at least 4 mounts to the DXI. Each mount was to one unique IP address.
mount o vers=3,tcp,rsize=1048576,wsize=1048576 NOTE: The older versions of Linux (pre-2.6.16) only support maximum of 32K rsize/wsizes that are not optimal. 2.6.18 and recent versions showed better performance.
Quantum Corporation
Page 15
Step # 1
To make them permanent, add the following lines into the /etc/rc.net file: /usr/sbin/no -o tcp_sendspace=4194304 /usr/sbin/no -o tcp_recvspace=4194304
Step# 1
Mount with rsize/wsize set to 1M.
mount o vers=3,tcp,rsize=1048576,wsize=1048576 NOTE: The AIX in the test ignored the rsize, wsize and defaulted to 65536
Quantum Corporation
Page 16
Step # 1
2 3
CIFS Configuration Options: Under Windows XP SP2, the "DefaultReceiveWindow" value in the AFD branch of the Registry takes precedence over the RWIN values in the TCP branch. Under SP1, it is the other way around, the values in the TCP branch override the AFD value and go out in TCP packet headers. Afd.sys is the kernel-mode driver that is used to support Windows Sockets applications. When there are three default values, the default is calculated based on the amount of memory detected in the system: The first value is the default for smaller computers (less than 19 MB). The second value is the default for medium computers (<32 MB on Windows XP Professional, <64 MB on Windows Server 2003). The third value is the default for large computers (>32 MB on Windows XP Professional, >64 MB on Windows Server 2003).
The following values should be set to 16384 on larger backup servers, or the defaults based on the machine sizes listed above.
HKEY_LOCAL_MACHINE \SYSTEM\CurrentControlSet\Services\Afd\Parameters DefaultReceiveWindow
Quantum Corporation
Page 17
The number of receive bytes that AFD buffers on a connection before imposing flow control. For some applications, a larger value here gives slightly better performance at the expense of increased resource utilization. Applications can modify this value on a per-socket basis with the SO_RCVBUF socket option.
DefaultSendWindow Value Type: REG_DWORD Default: 4096/8192/8192 This is similar to DefaultReceiveWindow, but for the send side of connections.
Step# 1
1. 2.
Go to Start>Run and type REGEDT32. Open the Registry Key: HKEY_LOCAL_MACHINE\SYSTEM\CURRENTCONTROLSET\ SERVICES\LANMANWORKSTATION\PARAMTERS\
3.
Locate the Key SESSTIMEOUT in the right panel. NOTE: If the SESSTIMEOUT key does not exist: Click in the right panel and select New and DWORD value. Create a new key, SESSTIMEOUT.
4.
Now all new Read append Operations will wait for 600 seconds before they Timeout. This gives the blockpool enough time to reconstruct the data out of the blockpool.
Quantum Corporation
Page 18
Microsoft provides comprehensive guidelines for configuration and tuning for networks and storage in the documents referenced below.
Windows XP
http://support.microsoft.com/kb/314053/
Windows 2003
http://download.microsoft.com/download/2/8/0/2800a518-7ac6-4aac-bd8574d2c52e1ec6/tuning.doc
Windows 2008
http://download.microsoft.com/download/b/b/5/bb50037f-e4ae-40d1-a8987cdfcf0ee9d8/WS08_STEP_BY_STEP_GUIDE/WS08PerformanceTuningGuide_En.doc x http://download.microsoft.com/download/9/c/5/9c5b2167-8017-4bae-9fded599bac8184a/Perf-tun-srv.docx
Windows Vista
http://download.microsoft.com/download/c/2/6/c26893a6-46c7-4b5c-b287830216597340/TCPIP_Reg.doc
Quantum Corporation
Page 19
Utilization Recommendations
Deduplication Limitations
Deduplication results can be negatively impacted by compression, encryption, software deduplication and, multiplexing. These functions all change the data stream in a way that obscures patterns in the data content that will reduce the performance and deduplication from any downstream appliance including DXi. To obtain effective deduplication rates, users should NOT encrypt, deduplicate, compress, or multiplex their backup data before sending it to a DXi appliance. The use of multiplexing was intended for slow source data and a minimum transfer rate required by physical tape drives. Multiplexing backup streams was intended to provide a more efficient use of a limited number of physical tape drives. Since the virtual tape drives in the DXi systems are not susceptible to performance losses from slow data transfer rates, the number of virtual tape drives can easily be increased in quantity without any time penalty for re-positioning. It is not necessary to use multiplexing with the DXi systems. Additionally, the multiplexing adds additional header information to the data lowering the deduplication ratio.
Space Reclamation
Space management involves two processes: data reconciliation and data reclamation. Data reconciliation is used to create a list of what can be removed. It runs automatically every twelve hours at noon and midnight, unless data reclamation is running. Data reclamation is the process of deleting the data on the data reconciliation list. It can be scheduled or run manually. There is significant overhead associated with this process, and therefore it should not be run during periods of high appliance use. Replication, reclamation, and backup stream ingest all consume system resources, and consequently should not all be done at the same time. Quantum recommends scheduling daily reconciliation and reclamation to manage the available space. The scheduled time should be configured to start the data reclamation process after the daily backups are complete. The default schedule is weekly, and the default time for the data reclamation is set to 12:00AM Sunday. These parameters are user configurable and should be configured for the customer backup window.
Quantum Corporation
Page 20
on the tape at the beginning of tape. This new label is a data block written to the virtual tape cartridge. The DXi-Series VTL will act similar to a physical tape and the data after the label becomes no longer accessible. The space reclamation can be initiated by the scheduled time or started manually from the DXiseries GUI. Additional information about the reclamation process can be found in the DXi-Series user manual. Note: Running the bplabel utility on a tape will make the contents inaccessible. Make sure the tape is expired before labeling the cartridge.
Additional Information
Quantum web site: http://www.quantum.com Symantec web site: http://www.symantec.com Microsoft web site: http://www.microsoft.com Guardian web site: http://www.quantum.com/ServiceandSupport/Services/GuardianInformation/index.aspx StorageCare Vision web site: http://www.quantum.com/Products/Software/Storagecarevision/Index.aspx Quantum Service web site: http://www.quantum.com/ServiceandSupport/Index.aspx Call Center Americas. To contact our world-class support representatives, please refer to the information below:
Telephone: 800-284-5101 (toll-free), 949-725-2100 (local/long distance) Hours of operation (subject to change without notice): 7 days a week, 24 hours a day with valid contract All other contracts can contact Quantum during normal business days from 5AM to 5PM US Pacific Time.
Quantum Corporation
Page 21