Planning preferred practices
This chapter describes the type of information to consider as you plan your IBM Scale Out Network Attached Storage (SONAS) solution implementation. It is important to consider critical solution requirements.
Record this information early in the SONAS solution planning process by using the SONAS survey questionnaire. This questionnaire is provided to you by your IBM representative to begin your cluster planning. This IBM Redbooks publication is meant to highlight the key points, but it does not describe all aspects of the questionnaire.
This chapter includes the following information:
1.1 Worksheet-oriented qualification and sizing review
When you are planning a new SONAS solution, be aware of some important considerations for the requirements of your current environment. These considerations include settings, expectations, needs, and common working habits. Considering all critical NAS solution requirements is key to a successful cluster planning project. With proper planning, you can avoid performance problems and functionality defects and build a system that meets
your needs.
your needs.
As a pre-sales activity, your IBM Account team presents you with a SONAS questionnaire, which is similar to the one in Figure 1-1. The information that you provide helps you and IBM to understand the important information and requirements of your particular solution. Take the time that is required to respond as completely and as accurately as possible to ensure that the cluster design is optimal.
Figure 1-1 SONAS and IBM Storwize V7000 Unified questionnaire
If your organization is considering SONAS and your IBM Sales representative has not reviewed the questionnaire with your staff, request that review as a component of preferred practice. It is better to do it late than not to do it, and it helps you to capture the full scope of your needs in a single worksheet.
The purpose of the solution preparation questionnaire is to record as much information as possible to help clarify the requirements for the environment on which SONAS is to be installed. The questionnaire also serves as a single point of reference to capture current usage patterns within your environment. It gives you an appreciation of what needs to be considered to facilitate a successful implementation and migration.
This chapter provides information about the following topics, which can help planners understand the key targets for solution planning:
1.1.1 Environment details
Unless this installation is a net-new service, you are asked to provide information about the current NAS storage environment that will be replaced. This information helps the IBM team accurately size, compare, and qualify an appropriate solution. This section describes many key points that are considered in a full NAS solution assessment.
|
Important: Always ensure that you are using the current survey that is available from your IBM Solution sales team.
|
The following information is included:
•Opportunity details
The evaluation begins by identifying source data collection resources and generally describing the goals of the project.
It is helpful to make sure that the people who provide the details for the solution planning survey can be identified. In some cases, the feedback details require further discussion with solution subject matter experts (SMEs). The questionnaire serves as the initial solution planning guide. Sizing and solution qualification and services are based on it.
|
Remember: Not every question applies directly to your needs. Answer as completely and accurately as you can. Select only the boxes for protocols that you need explicitly or immediately. If you might use options in the future, describe those requirements in the narrative explanation boxes that are provided throughout the survey.
|
•Capacity, growth rates, and storage tier planning
Always express capacities in real sizes. Declare capacity requirements without compression or data deduplication. Space reduction features are often not reliable enough to predict explicit capacity goals. Data profiles must be considered explicitly to see value from space reduction techniques.
Growth rates are often difficult to predict, but try to provide an expectation for one-year growth at the minimum. One value of the SONAS product is the ability to scale non-disruptively to very high capacities.
Storage tiers are a good way to segment data storage behind SONAS or Storwize V7000 Unified systems. Try to predict your tiering needs with the growth rate in your planning.
|
Remember: The typical sizing for metadata is roughly 3% - 5%, depending on whether the metadata is replicated.
|
•Tape-based storage tiers
Hierarchical storage management (HSM) space management is a popular archive solution that can be managed simply with efficient HSM and information lifecycle management (ILM) policies. However, as this tape-drive-managed access is passed via Transmission Control Protocol/Internet Protocol (TCP/IP) through the cluster and the IBM Tivoli Storage Manager server, there can be high latencies that are associated with file access and data recall (from tape back to disk).
For this reason, it is critical to plan targeting data that is not likely to be recalled in a large scale from these archives. Retrieval of a few files here or there does not have a major effect on the system. However, retrieving millions of files at a time can create resource challenges for the SONAS and for the correlating Tivoli Storage Manager environment.
|
Tip: Plan use for HSM tiers for true archive, and not an inexpensive tier of storage for use with data that can change or be subject to frequent recall.
|
1.1.2 Performance requirements
This section describes some elements of the Performance Requirements section of the questionnaire, as shown in Figure 1-2.
Figure 1-2 Image of Performance Requirements section of the questionnaire
When you are planning a new SONAS solution, remember the following key performance requirements:
•How performance is monitored in current NAS environments
If this installation is a net-new NAS environment, you might not have a plan for performance monitoring yet. However, in many cases, clients do have existing solutions that they are replacing, for one reason or another. In this case, provide the performance metrics from those environments to clearly map the new system configurations.
|
Tip: There is little value in providing solution sizing requirements that are based solely on advertised “high watermark” specifications. Typically, this method adds cost to solutions by providing functions that clients probably will not use. There is no way to size for realistic performance goals when data and input/output (I/O) workload characteristics are not clearly understood. Small file random workloads perform differently than large file sequential workloads.
|
•Workload descriptions
Take the time to explain the workload characteristics that you do and do not know. For example, explain whether it is primarily for random or sequential access. Specify whether there are mostly large or small files. Explain whether there are peaks and valleys in client use. Specify whether files are reread often (cache hit ratios). That information helps with sizing the front and back-end storage requirements, and helps the team that is designing the implementation to prepare to meet the performance goals at the lowest possible cost.
•Description of your storage needs (in your own words)
Space is provided for you to elaborate the workload and performance characteristics. It is provided in case the typical questions do not exactly cover the requirements.
1.1.3 Client applications and use cases
This section describes the planned use case for the network-attached storage (NAS). Figure 1-3 shows the Client and Application section of the survey.
Figure 1-3 Image of Client and Application section of the questionnaire
Be prepared to consider the following topics:
•Specify whether you plan to use the NAS solution for VMware storage. If so, describe what things you plan to support on the shares. If it is not otherwise obvious, describe the planned predominant use case for the solution.
•The number of concurrent users and home directory use cases might require special planning or extra interface nodes to manage the number of concurrent active users.
|
Important: SONAS can support multiple use cases. However, one size does not fit all. In some cases, using separate file systems or network groups makes sense for providing a more multi-tenant approach to I/O isolation. Early planning for this situation prevents surprises and inevitable scaling.
|
1.1.4 Client communications protocols
Client connection speeds, TCP/IP and port requirements, and specific protocol requirements are required to qualify SONAS as a solution. Figure 1-4 shows a sample of the Client Communication Protocols section of the questionnaire.
Figure 1-4 Image of the Client Communications Protocol section of the questionnaire
SONAS is developed to meet the requirements that clients request, and the roadmap (for development) has largely followed that focus of evolution. IBM delivers a SONAS implementation only when it can be delivered well.
In some cases, advanced solutions are approved on a per-client, case-by-case basis, using a request for price quotation (RPQ) to ensure that client deployments are clearly configured and therefore supported.
There is little value in becoming dependent on protocol models that are still in a state of flux (from a standards development perspective) or are not ready from a client stability perspective. For this reason, the questionnaire asks that you remain honest about protocols you need for day one and describe the protocols that you are interested in using in the future.
If you are not running Server Message Block 3 (SMB3) or Network File System version 4 (NFSv4) today, but plan to in the future, be sure to make that plan clear in your requirements summary.
|
Remember: Some protocols and check boxes in the survey might refer to points of interest that are not in the current release of SONAS or Storwize V7000 Unified, but are available in other forms of IBM-supported NAS solutions. If you want to use them in your solution, be clear about when you need them, why, and how, so that the most appropriate solution for your business requirements and applications is provided.
|
1.1.5 Authentication requirements
Figure 1-5 shows the Authentication Requirements section of the questionnaire.
Figure 1-5 Image of the Authentication Section of the questionnaire
The preferred practice is generally to use Active Directory and Services for UNIX (AD + SFU) or Lightweight Directory Access Protocol (LDAP) for authentication. However, several forms of authentication are supported.
|
Remember: Authentication and SONAS need expanded visibility and cross-team understanding. This survey asks the questions that help to qualify the solutions that fit your business needs. In many situations, migrations from existing platforms also need consolidation of authentication along with storage.
In this case, IBM can help you design solutions that simplify the long-term goals of your enterprise storage. In any case, ensure that your authentication personnel clearly communicate with solution authentication experts to understand authentication requirements and configurations.
|
1.1.6 NFS protocol requirements
NFS and other protocol requirements are gathered as part of the questionnaire. This section is shown in Figure 1-6.
Figure 1-6 Image of the NFS and Other Protocol Requirements section of the questionnaire
For more information about supported NFS versions and file systems, see the IBM SONAS Support Matrix in the IBM Knowledge Center:
1.1.7 Data protection
Data protection covers several features and options with scale-out NAS products. The questionnaire includes a short list of questions to consider to help you plan for implementing a SONAS or Storwize V7000 Unified solution. Figure 1-7 shows the Data Protection section of the questionnaire.
Figure 1-7 Image of the Data Protection section of the questionnaire
Snapshots are a common form of protecting data from unwanted changes, and allowing customer-recoverable snapshots is common. The questionnaire is designed to show your planned use of snapshots. The remaining chapters in this Redbooks publication include extensive information about snapshot management.
This book also refers to preferred practice information for asynchronous replication. A clear understanding of replication goals is critical for advanced sizing and performance planning. For more information about replication, see Chapter 7, “Data protection” on page 209 and the Planning for data protection topic in the IBM Knowledge Center:
Also see the IBM Active Cloud Engine® information in IBM SONAS Implementation Guide, SG24-7962.
Data backup strategies are the obvious consideration for data protection. SONAS and Storwize V7000 Unified have built-in integration for Tivoli Storage Manager-based backup solutions or Network Data Management Protocol (NDMP). The questionnaire challenges the design team to consider the solution backup requirements along with the sizing information.
In many cases, this is one aspect of your environment that will grow along with your scale-out NAS. Although it is not typically installed for the explicit purpose of disaster recovery (DR), it is extremely useful for aspects of data protection and file-based restore.
Because there are many components to consider in the planning of the SONAS solution implementation, it is a preferred practice to plan this thoroughly, along with your backup expansion plan, when planning your scale out solution.
1.1.8 Data center requirements
The data center requirements can be important for many reasons. You need to identify issues that might affect installation, power, cooling, racking, delivery, or ongoing maintenance from IBM service and support resources. The Data Center Requirements section of the questionnaire is shown in Figure 1-8.
Figure 1-8 Image of the Data Center requirements section of the questionnaire
Be sure to list all hardware or data center access concerns in the narrative box that is provided in the questionnaire.
1.1.9 Service requirements
The services section is a place to list any services with which you might need assistance in any phase of your SONAS solution implementation project. You can also list preliminary services that might help you prepare and make decisions, such as knowledge transfer for advanced skills, proof of concepts, cluster installation, or data migration.
Narrative boxes are provided for free-form explanations of what you want to discuss or request for services requirements. The questionnaire leads to additional discussions with subject matter experts. Brief explanations typically suffice in this section of the questionnaire. See Figure 1-9.
Figure 1-9 The Service Requirements section of the questionnaire
1.2 Staff and training
SONAS and Storwize V7000 Unified are simple platforms to manage from a graphical user interface (GUI) and command-line interface (CLI) perspective. However, SONAS can be a complex environment that involves many external variables.
Knowledge of the following topics is critical for supporting a SONAS environment:
•Networking
•Block storage
•Network file share protocols, such as Network File System (NFS), Common Internet File System (CIFS), and Hypertext Transfer Protocol (HTTP)
•File access control lists (ACLs), and ACL management practices
•Network share exports and options that affect shares and share access controls
•Client platforms and how clients tune, manage, and control share access
•Client platform security
•Data Access Authentication
•Data backup and Restore
•Domain name server (DNS)
•Network Time Protocol (NTP) general troubleshooting skills for all of the preceding items
These topics are critical skills for NAS administration. Also consider requirements to work across teams; to articulate, document, and communicate issue descriptions; troubleshoot processes and progress; and manage advanced NAS topics.
SONAS implementations require many skills and can affect personnel in many roles. Staffing the SONAS or Storwize V7000 Unified solution is of critical importance to success as you plan a SONAS project. Plan training, backup resources, or staff augmentation to address skills gaps.
Maintaining detailed run books is a common preferred practice. These books can be used to provide instructional guidance and process management. Proper change control is critical to tracking events and issue resolution for all pertinent complex SONAS tasks.
1.2.1 Staffing considerations
This section describes requirements for SONAS staffing resources. In in some cases, multiple roles can be shared by qualified resources.
NAS management
This resource is a dedicated resource to decide, track, manage, and communicate all status, requirements, issues, and resources that are assigned to SONAS. This resource is also responsible for scale out purchase decisions.
NAS project manager
This resource is a dedicated resource to document cluster requirements across internal teams and external vendors. This resource organizes project schedules resource requirements and event timing. They also serve as a point of contact for recording and organizing all team event management and coordination of resources that are required to establish NAS policy, protocol, and maintenance.
NAS technical lead
This resource is responsible for the technical skills that are required in all stations of support. They are responsible for the accurate capture of all aspects of cluster design and layout, and can communicate up and down on all aspects of the community of NAS resource issues and requirements at an in-depth technical level for all the preceding points.
This resource must be able to work across teams with collaboration, and possess analytical expertise to help solve and coordinate all efforts that are related to scale out NAS as deployed in the client environment.
Network resource
This resource is an advanced-level network engineer who is assigned to client networking technology that can be used when network issues are causing problems. This resource must be able to articulate, communicate, and document across teams. These skills are used to help manage and troubleshoot issues with network-attached storage for all server and client networks, openly and constructively as needed.
Authentication resource
This resource is an advanced-level authentication expert who is familiar with all client data authentication environments used for SONAS. This resource must be familiar with and able to articulate, document, and communicate all aspects of client and SONAS authentication requirements and assist with all authentication challenges and troubleshooting as needed.
Backup and restore resource
This resource must understand all technical aspects of the backup and replication solution in place for SONAS. This resource must also understand all disaster recovery (DR) or business continuance requirements of the client, and how that process can be managed, validated, and implemented when necessary. This resource must know and help articulate all requirements for both the SONAS and the backup service in scale with solution requirements.
Client platform specialists
For any platform that is supported in the client SONAS environment, the client should have a technical expert that can help verify client requirements for access and control of shared data on those platforms. The expert should also verify patch, driver, and operating-system level requirements for client platform readiness, security, and data integrity in the specified platforms.
Client application experts
Each advanced application that uses shared data from the SONAS solution must include a representative application expert. This expert can help articulate, document, and communicate application requirements, conditions, status, and troubleshooting for NAS service, as needed.
1.2.2 Education
The Scale Out Network Attached Storage (SONAS) technical training course is available in the IBM Training catalog. Search for SONAS on the following website:
This course covers details and usage of the SONAS product. Hands-on exercises include configuration aspects of the product using the CLI and the GUI. The participants create shares on the SONAS and verify file access from Windows and Linux clients.
They also learn how to configure the SONAS storage and file systems, back up with Tivoli Storage Manager, automated storage tiering with ILM and HSM, along with asynchronous replication between two SONAS systems.
1.3 Environment
It is important to prepare the environment for your SONAS before delivery and deployment of the solution. This section describes the environmental requirements for your data center:
•Space requirements
•Power requirements
•Cooling requirements
•Network requirements
1.3.1 Space requirements
Space clearance is important. When you prepare the environment for the SONAS, consider the distance limitations between nodes. The nodes connect to each other by InfiniBand cables through InfiniBand switches. The location of the SONAS frames and the distance between them is limited due to InfiniBand technology.
Another consideration is future growth. Remember that, for future growth and expansion, you need to plan space close to the base frame. A typical configuration is one row for your SONAS hardware, but the main concern is to keep the hardware in a concentrated location.
Your location must meet floor weight load requirements. You can find detailed information about floor load requirements, rack dimensions, and individual weights in the IBM SONAS Implementation Guide, SG24-7962 publication.
1.3.2 Power requirements
Ensure that the SONAS environment meets the proper ac power and voltage requirements.
SONAS has redundant construction for power. Every frame uses two primary and two secondary power cords (four per rack). You must plan the power environment so that one circuit can handle the full load. The full consumption of the SONAS frame especially depends on the populated nodes and the types of disk drives, in case of the storage pod, as shown in Figure 1-10, Figure 1-11, and Figure 1-12. 15,000 revolutions per minute (RPM) or 10,000 RPM SAS drives use more power than the Near-Line SAS drives.
Figure 1-10 Power use for SONAS frames
Figure 1-11 Power use for SONAS storage pods
Figure 1-12 Power use for SONAS individual components
Measured power has many variables that can change the value of the power usage, and it needs to be used and interpreted with caution. Depending on what input voltage is used in your lab, what data I/O is running, and your lab infrastructure, all can play a role on measured data results. So, it is imperative to note that the measured data is an estimate, and is subject to change dependent on hardware upgrades, lab setup, and I/O choice.
If the hardware that is used is upgraded or the technology changes, the measured power data changes. Measured data is a good reference point, but set up your lab to accommodate the maximum rack power consumption numbers. With the 30A power cords, each rack can draw 9600 W. With 60A power cords, each rack can draw 19200 W. Measured data is not published.
|
Important: The whole redundant architecture is ineffective if the SONAS is not connected to a separate power circuit. In special cases, if you require a high-availability (HA) solution, you must provide a separate power circuit from a different power provider.
|
Redundancy and power distribution
Redundancy might not be intuitive. Ensure that your power specialists have a complete understanding of the wired redundancy plan for your SONAS implementation.
SONAS power distribution
The SONAS RXA cabinet has four power-distribution units that help distribute the redundancy of power from all components in the SONAS (see Figure 1-13).
Each rack has either four intelligent power-distribution units (iPDUs) or four base PDUs. The iPDUs collect energy use information from energy-management components in IBM devices, and report the data to the Active Energy Manager feature of IBM Systems Director, if it is installed on a customer server. IBM Systems Director can measure and monitor power consumption.
Figure 1-13 Power Distribution Unit (PDU) in a SONAS RXA frame
Each rack requires four power cords, or two features. Each power cord feature is two cords.
Four power-distribution units (PDUs) in an IBM SONAS rack each require a separate power source. Each of the PDUs contains twelve 200 - 240 V ac outlets that provide power to the drawers and devices in the rack, as shown in Figure 1-14 on page 16.
|
Note: All four PDUs in the rack must be configured.
|
The IBM SONAS PDUs are split vertically in the rack with upper two PDUs as the secondary PDUs and lower two as the primary PDUs. To provide power redundancy to all the components in the rack, all four PDUs must be plugged in.
If either the upper two (secondary) PDUs or the lower two (primary) PDUs have power, the rack can still function, but without redundancy. However, if only one of the upper PDUs and one of the lower PDUs have power, that is, if only the right two PDUs, or the left two PDUs have power, the rack cannot function and some of the components in the rack will be
without power.
without power.
You need to plug the upper two (secondary) PDUs into one power bus and the lower two (primary) PDUs into another power bus. So, even if the power to one bus is lost or disconnected, the rack still has power. If you connect the two right PDUs to one bus and the two left PDUs to another bus (splitting the power horizontally), and if one bus goes down or is disconnected, the rack cannot function because some components will not have power. See Figure 1-14.
Figure 1-14 Power redundancy (graphical review)
The power redundancy layout is important to understand before you connect your SONAS frames to the data center Uninterruptible Power Supply Units (UPSs), and redundant power source feeds.
Cooling requirements
For the best performance and proper operation, you must optimize the cooling system of your SONAS storage solution. Use a raised floor to increase air circulation, in combination with perforated tiles. The air flow enters at the front of the rack and leaves at the back as shown in Figure 1-15.
Figure 1-15 Air flow
To prevent the air that is leaving the rack from entering the intake of another piece of equipment, racks should be positioned in alternating rows, back-to-back and front-to-front, as shown in Figure 1-16. The front of racks should be positioned on floor-tile seams, with a full line of perforated tiles immediately in front of the racks, and the air temperature in front of the rack at less than 27 degrees C.
Figure 1-16 Rack positioning for proper airflow
Often forgotten in an installation, rack filler panels are an important addition. Blank panels cover up unused rack space and prevent unwanted access to equipment. Blank panels are also crucial to thermal management by controlling and restricting airflow through a rack enclosure. All unused rack space requires the usage of a filler panel.
Figure 1-17 shows the temperature and humidity information.
Figure 1-17 Temperature and humidity information
SONAS hardware that runs continuously must be within the recommended operating environment. Operation of the SONAS hardware at the maximum allowable temperature is only intended for short durations such as can occur during a hard drive or power supply replacement. Continuous operation above the recommended maximum temperature increases the probability of component failure.
For more information about this type of configuration, see the SONAS Introduction and Planning Guide, GA32-0716.
1.3.3 Network requirements
SONAS means Scale Out Network Attached Storage, and the network preparation and configuration are important for overall performance. For fast and correct implementation, follow these basic principles:
•Network administration
In most cases, the network and storage administration are separated. Storage configurations are internal to SONAS. However, you must deploy SONAS external networking into the current client network environment. This deployment requires a careful review of all network requirements:
– VLAN configurations
– Switch configurations
For an active-active network connection (as previously described) you must set up current switches with appropriate settings.
– Firewall configurations
– Free IP addresses
The planning guide excerpt in Table 1-1 helps to identify your networking plan for SONAS interface nodes. For example, in many cases, assigning multiple IP addresses to each interface node helps to distribute workloads evenly across the remaining interface nodes if one node fails.
Table 1-1 Interface node planning information
|
Number of Interface nodes
|
Number of bonds per Interface node
|
Number of IP addresses for management
|
Number of IP addresses for client connections
|
Summary of IP addresses
|
|
2
|
1
|
3
|
1
|
4
|
|
2
|
2
|
5
|
||
|
3
|
3
|
6
|
||
|
4
|
4
|
7
|
||
|
3
|
1
|
3
|
6
|
|
|
2
|
6
|
9
|
||
|
3
|
9
|
12
|
||
|
4
|
12
|
15
|
||
|
n
|
k
|
3
|
(n - 1) * k
|
(n - 1) * k + 3
|
|
Note: Every frame is configured with two mandatory, 50-port SMC gigabit Ethernet (GbE) switches that are used for the internal network. Every node, Interface (2851-SI1), Management (2851-SM1), and Storage (2851-SS1), is connected to the SMC switches with three separate connections, a primary, secondary, and maintenance Ethernet cable.
The InfiniBand switches and the iPDUs also have an Ethernet connection to these SMC switches. The disk storage units are not connected to the SMC switches. Rather, each Redundant Array of Independent Disks (RAID) controller is connected to each of the storage nodes, or, in the case of gateway attached storage, they are managed separately from the SONAS internal network infrastructure.
|
1.4 Sizing
Sizing is the most important component for successful solution planning. Conducting a complete and thorough review of the current environment, including all requirements and expectations, along with a full Technical Delivery Assurance (TDA) review, helps to close any gaps. The basic task is to determine what is the most appropriate SONAS configuration for your needs.
The purpose of the primary exercise is to avoid over sizing (to save on cost), but also to avoid under sizing. An undersized system might fail early on required capacity or performance and therefore require scaling growth early in the implementation cycle. The best way to begin answering the tough questions is to fully understand your client and application behavior. This includes I/O and access pattern requirements.
IBM offers advanced help in this pre-sales evaluation process. So, it is common to expect a complete review with some interaction with component experts while you size your solution for production use.
For detailed information about SONAS solution sizing, see IBM Scale Out Network Attached Storage Introduction and Planning Guide, GA32-0716 and the SONAS planning information in the IBM Knowledge Center:
The following section provides several high-level, typical sizing considerations to help to provide an overview of topic considerations. Here are some key points for a successful sizing.
Capacity sizing
The capacity sizing question is probably the easiest question to answer. The most important thing is not to build it at 100% utilization of capacity, because that choice restricts the growth availability within the solution that is purchased. Take time to consider the effect of snapshots, snapshot retention requirements, and file clones.
Make sure to plan for expanded file capacity if the data you plan to migrate is stored in tape archives, or compressed or de-duplicated file stores. Calculate enough capacity for the early growth. It might be a good starting point if you understand the needed data capacity and add 30 - 40%.
Storage tiers
The type of disks, number of spindles, and the RAID configuration can dramatically affect performance and capacity in addition to the cost of the overall solution.
Solid-state drives (SSDs) offer transaction speed benefits but also cost the most (financially). For this reason, it might make complete sense to put metadata on SSDs, while you plan the tier 1 data on high-speed serial-attached SCSI (SAS) and tier 2 data on Near-Line SAS (NLSAS), and perhaps even plan for some long-term archive to a tape-based tier. Discuss and evaluate this planning before you make your initial purchase.
Back-end sizing
Each storage node pair can drive up to about 3 gigabytes per second (GBps) in aggregate performance today, with peaks that spike well above that. Some clients drive that range of bandwidth. Note however, that this estimate is a typical estimate and I/O and workload patterns can deviate from that estimate). Consider the most common mixed workload performance to average 1.5 - 2 GBps per storage node pair.
This means that if the back-end storage configurations are inferior to supporting that performance, your number will be lower, and if the back-end storage is twice that fast, it will not likely be fully used through the storage node pair workload ceilings (today). So, planning back-end storage behind each storage node pair to enable using sequential performance somewhere between 3 - 4 GBps is a reasonable target for maximizing the storage channels.
Front-end sizing
Each interface node can drive up to 3.5 GBps from cache, with 4 x 10 GbE teamed converged network adapter (CNA) ports. However, do not expect all your I/O to come from cache. The pass-through expectation for each interface node with optimal networking remains somewhere in the vicinity of approximately 2 - 2.5 GBps for mixed loads (with about a 15% - 20% data reuse rate) with multiple 10 GbE ports teamed.
You might also find a performance limit of concurrent user activity for each interface node. What is meant by that is that there is a soft limit of approximately 2500 - 3000 active concurrent users per interface node today. Where users do often behave differently in many circumstances, you might consider this a typical pattern for common client use cases with concurrent, active users per interface node.
Therefore, if you expected to have 7500 active concurrent users constantly using your data shares, you might want three interface nodes in your environment to support that activity. To provide adequate protection of performance capacity if an interface node fails, you might even consider having four (for added redundancy). This limit is a soft limit because it depends on how active the users are.
Performance sizing
This is the most complex part of the sizing. You must collect much information for a fully appreciated performance sizing exercise. However, always remember that the nature of scale out NAS is that it (SONAS) enables you to grow the environment when expansion is required. So, when you do approach the ceiling or a bottleneck, evaluate where it is carefully, and expand your cluster to move that mark.
Sizing workflow
Figure 1-18 on page 23 shows a high-level sizing workflow to help you understand the phases of sizing.
|
Note: Sizing seems complex, but scientific review of planned requirements is important. SONAS solutions include subject matter expert (SME) review with every purchase. If you feel that this review has not been offered, ask your storage representative to speak to a storage performance expert. IBM can help size storage solution requirements.
|
Figure 1-18 is a diagram to help you appreciate some of the considerations in review sizing requirements for scale out NAS.
Figure 1-18 Sizing workflow
..................Content has been hidden....................
You can't read the all page of ebook, please click here login for view all page.