Optimize Desktop Virtualization via Highly
Manageable Scale Out Storage Clouds
Da
ta
Center
SAN
Infrastructure
openBench Labs
Author: Jack Fegreus, Ph.D. Managing Director openBench Labs http://www.openBench.com September 22, 2010
Optimize Desktop Virtualization via Highly
Manageable Scale Out Storage Clouds
Analysis
:
Jack Fegreus is Managing Director of openBench Labs and consults through Ridgetop Research. He also contributes to InfoStor, Virtual Strategy Magazine, and Open Magazine, and serves as CTO of Strategic Communications. Previously he was Editor in Chief of Open Magazine, Data Storage, BackOffice CTO, Client/Server Today, and Digital Review. Jack also served as a consultant to Demax Software and was IT Director at Riley Stoker Corp. Jack holds a Ph.D. in Mathematics and worked on the application of computers to symbolic logic.
Table of Contents
Executive Summary
04
VDI Test Scenario
07
VDI Automation & ROI
12
Customer Value
16
03
DISTRIBUTEDCOMPUTING2.0
Today, Forrester Research pegs the typical SMB IT organization as having two years of basic virtualization experience. Often based on DAS, rather than SAN, storage, these initial projects have proved to be successful in generating savings through hardware consolidation. For these sites, the next phase in adoption will focus on improving manageability and flexibility, which requires the adoption of shared storage and often requires upgrades to network and server infrastructure.
The most dramatic changes, however, are yet to come. The real revolution in distributed computing starts with the adoption of hosted virtual desktops. In a hosted desktop scenario, IT centralizes the running of PC operating systems and applications on virtual machines (VMs) hosted on servers in the data center while streaming a display protocol to a light-weight client application that can be run on various devices on the network.
Desktop virtualization builds on the notions of server
virtualization, which utilizes a physical server to host multiple virtual machines that run their own server-class OS and applications, which are typically in the form of backend services. In a desktop virtualization scenario, however, the hosted desktop VMs need to support a highly GUI-centric environment running interactive client applications. As a result, a virtual desktop infrastructure (VDI) introduces a
Executive Summary
Executive Summary
“F
or sites implementing a VOE, Xiotech provides Virtual View, a
software module for ISE™ Manager, which integrates with VMware,
Hyper-V, and Citrix Xen hosts and allows IT administrators to centrally
configure and manage storage resources on every VM.”
04
openBench Labs Test Briefing:
Xiotech
®Emprise
™5000 ISE
™-based Storage
1) Minimize Storage Management Costs:Host-based ISE Manager software aggregates the information from each ISE storage device into a virtual ISE cloud for centralized administration, which can be extended to include the ability to drill down and manage the VMs running on a VMware, Hyper-V, or Citrix Xen server provisioned with storage from an ISE in the cloud.
2) Maximize Performance Automatically:Advanced firmware stripes data at the disk head level, which eliminates the need for IT administrators to create disk-centric storage pools and creates a low-latency environment that facilitates the IO characteristics of a Virtual Desktop Infrastructure.
Iometer Streaming I/O Benchmark:For a VM running Windows 7, streaming read I/O throughput averaged 370MB per second and streaming write data average 175MB per second.
Iometer I/O Operations Benchmark:With the low latency of the Emprise 5000, 8KB reads and writes (80/20 percent mix), on a single VM sustained 3,500 IOPS with an average access time under 25ms, while two VMs sustained an I/O load of 6,500 IOPS.
3) Improve I/O Scalability: Storage capacity and performance can be easily expanded by adding multiple Emprise 5000 systems. This adds local cache and processing power via active-active Managed Reliability Controllers, which locally manage I/O processes.
4) Improve Storage Reliability: Intelligent Storage Element (ISE) technology provides autonomic self-healing to minimize storage-based service events.
number of new specialized technologies—from session brokering to display streaming— for IT administrators to support.
Centralizing virtual PC operations in a datacenter with hosted desktop VMs that have applications and data residing securely on servers in the datacenter makes managing and securing desktops significantly easier for IT. This breaks the problem remediation model of a traditional PC environment, by ending the need to attend to dispersed physical clients individually. To fully leverage such a scheme, however, even more advanced VDI solutions are needed to address such issues as virtual desktop provisioning and the streaming of PC applications and disk images.
The potential impact of a VDI environment is staggering. In terms of today’s $150 billion worldwide market in business PCs, Gartner pegs client systems deployed on VMs to be only around 500,000—not much more than a rounding error. Nonetheless, as IT finds it conceptually easy to leverage existing infrastructure to offset VDI entry costs, Gartner projects the percent of new business PCs being deployed on VMs to rapidly rise to 40%. According to Gartner, IT in the US will lead this trend by migrating 30 percent of their installed base of desktop PCs to VMs by 2014. At that rate, the ranks of VMs running client systems will swell to over 18 million.
That rosy scenario for VDI is not without a dark side. IDG surveyed CIOs
implementing server virtualization and found that the percent of CIOs experiencing an increase in the complexity of datacenter management rose to 67% from 47 percent at the end of 2008. That increase in perceived complexity raises a serious red flag for VDI, as best practices call for deploying desktop VMs four to eight times more densely than server VMs. What makes dense deployment plausible is the sporadic nature of desktop PC usage. While dense VM deployment enhances the potential for significant cost savings, dense deployment also increases the need for IT to be prepared for resource-utilization storms involving I/O, memory, and CPU resources.
Whether to advance the manageability and flexibility of server virtualization or make the transition from server to desktop virtualization, a SAN-based storage infrastructure that can scale out in capacity and performance is pivotal for IT to maximize the return on investment (ROI) of any virtualization or consolidation initiative. Storage resources are inextricably linked to the capital and operational expenses that IT must restructure continue to advance virtualization initiatives.
There are numerous formulas that rely on array controllers and disk spindles to ensure reliability and throughput for DAS storage built on the traditional “Just a Bunch of Disks (JBOD)” model. In a virtual operating environment (VOE), however,
hypervisors on host servers concentrate and randomize I/O from multiple VMs making a hash of traditional read-ahead and caching algorithms and creating an absolute requirement for hardware with minimal I/O latency. As a result, every Emprise 5000 added to a SAN fabric provides more cache and I/O processing power to the SAN.
Xiotech redefines the notion of a storage building block with a radically different
05
construct from traditional JBOD and RAID storage. Dubbed ISE™ technology, Xiotech utilizes sealed multi-drive DataPacs, which contain specially matched Seagate Fibre Channel (FC) drives with specialized firmware that provides detailed information about internal disk structures. Using that detailed disk structure knowledge, DataPacs stripe data at the drive-head level. IT administrators never create or reformat a RAID array. With an ISE, RAID decisions are reduced to a simple virtualized RAID 5 or RAID 10 check-off characteristic, which is assigned to a virtual volume as it is created. Gone are all I/O bandwidth issues associated with ports, controllers, spindles and RAID levels.
For IT’s bottom line, the most powerful impact of ISE technology, however, comes in the form of autonomic self-healing storage that is service avoidant. Xiotech matches the specialized drive firmware used in a DataPac with specialized firmware on Managed Reliability Controllers (MRCs), which have the ability to monitor and reduce the rate at which DataPac components fail by repairing many component failures in-situ and mitigating the impact of any failures that cannot be repaired. MRC remedial reconditioning extends to remanufacturing disks through head sparing and
depopulation, reformatting low-level track data, and rewriting servo and data tracks. To further simplify SAN management, an ISE isolates the function of its two external FC ports from that of its two internal MRCs. No notion of MRC array ownership is exposed on the SAN. This allows the ISE to maximize external SAN fabric traffic and internal data access independently. All host servers are provided with a view of the ISE as having two active MPIO ports ready to accept FC data frames. As a result, a typical server configuration with two FC ports has four paths for balancing fabric traffic.
Internally, the ISE optimally redirects the balanced read and write I/O requests arriving at the FC ports to the two MRCs in order to maximize internal throughput for the DataPacs. As a result, a Xiotech ISE slashes operating costs by eliminating all of the device-management tasks that IT administrators normally perform to maintain and optimize the performance of standard JBOD-based storage resources. In addition, the ISE’s heal-in-place technology allows the Emprise 5000 to reach reliability levels that are impossible for standard storage arrays. As a result, Xiotech is able to provide IT and OEM users with a five-year warranty that eliminates storage service renewal costs.
To further simplify storage-resource scale out and lower IT operating costs, Xiotech provides host-based software dubbed, ISE Manager, which aggregates all local
information resident at each ISE into a single virtual ISE cloud. More importunately for sites implementing a VOE, Xiotech provides Virtual View, a software module for ISE Manager, which integrates with VMware, Hyper-V, and Citrix Xen hosts and allows IT administrators to centrally configure and manage storage resources on every VM.
Executive Summary
PROVISIONING FORHIGHI/O THROUGHPUT
With server
virtualization rated the best way to optimize resource utilization and minimize the costs of IT operations, many SMB sites now typically run four to eight server VMs on a host server in production environments. As a result, each host must be capable of delivering a very high I/O throughput load via a limited number of physical connections.
To provide a VOE foundation, we used two servers running MS Windows 2008 Server R2 with Hyper-V
virtualization to host our virtual desktop
infrastructure. Each host server was equipped with a two quad-core CPUs, 8GB RAM, and a dual-port QLogic QLE2462 4Gbps Fibre Channel HBA. With the exception of a local disk for the OS, all storage on each Hyper-V VOE host was provisioned from a single
07
VDI Test Scenario
VDI Test Scenario
“T
o simplify the provisioning of VOE hosts and the VMs running on
those hosts, Xiotech provides host-based software—dubbed ISE
Manager—that aggregates information from each ISE-based storage
device into a virtual ISE cloud.”
Xiotech Emprise 5000.
Within our Hyper-V environment, we conservatively hosted 5 server VMs to support a VDI infrastructure created with Citrix® XenDesktop™ Enterprise Edition. In particular, we used three VMs running 32-bit implementations of Windows Server 2003 to support three Citrix functions: The Desktop Delivery Controller (DDC), which brokers the assignment of desktop VMs to users, the Citrix License Server, and the Provisioning Service (PvS), which streams a single OS disk image to multiple VMs.
A fourth VM running Windows Server 2003 supported a domain controller for our VDI environment. In addition to Active Directory, this server provided DHCP and DNS services to all VMs in the XenVDI domain. We used a fifth VM running Windows 2008 R2 to host MS System Center for Virtual Machine Manager (SCVMM). Finally, we used this VM server infrastructure to support a collection of ten desktop VMs running Windows 7 Enterprise with either a dedicated system disk or a system disk that was streamed from the PvS server.
VOE STORAGESCALEOUT
To anchor the FC SAN fabric in our VDI testing scenario, we utilized a Xiotech Emprise 5000 system with two “Balanced” DataPacs. Xiotech provides DataPacs tuned to support
transaction processing, data archiving, or multi-purpose—balanced— computing.
The two balanced DataPacs yielded a combined raw storage capacity for our Emprise 5000 of 8.7TB. We were able to provision that storage in a mix of RAID-5 and RAID 10 logical volumes, which determines the actual volume of storage deployed at host systems.
Traditional core-driven SAN fabrics are characterized by a large number of physical servers and a small number of storage devices. A VOE, however, presents a very different SAN topology. In a typical VOE scenario, lightly loaded servers are converted into VMs
08
VDI Test Scenario
HYPER-V STORAGERESOURCES
Using the Web GUI on our ISE, we provisioned a 1TB volume to store VMs and a 500GB volume for use on each of our Hyper-V hosts. In addition we provisioned an 800GB volume on our host, Dell1900VMb, to provide storage for a domain-wide library that would be created on a VM running MS System Center Virtual Machine Manager.
and consolidated on a small number of VOE hosts. In turn, the small collection of host servers drive the I/O load within the SAN creating an edge-driven topology.
More importantly, IT Service Level Agreements (SLAs) for business applications running on servers that have been converted into VMs must now take into account the interactions of multiple consolidated VMs. Even more vexing for IT is the problem presented by multiple VMs fluidly moving within a group of VOE hosts in order to balance CPU, memory, and I/O loads. This situation is further exacerbated in a VDI scenario as the density of desktop VMs deployed on host servers is invariably several times greater than the density of server VMs,
To simplify the
provisioning of VOE hosts and the VMs running on those hosts, Xiotech provides host-based software—dubbed ISE Manager—that aggregates information from each ISE-based storage device into a virtual ISE cloud. This architecture directly enhances manageability in a scale-out environment. What’s more, the Virtual View module for ISE Manager provides significant operational savings for IT
administrators in any VOE with a single Emprise 5000.
Virtual View for ISE Manager integrates with the host servers and management consoles in a VOE running VMware, Hyper-V and Citrix Xen. As a result, Virtual View provides IT administrators with the ability to manage all ISE-based volumes imported by a VOE host and then drill down on the VMs supported by that host. In this way, IT administrators garner a single-pane-of-glass interface through which they can configure and manage storage resources for both VOE hosts and their resident VMs.
BREAKINGPERFORMANCEBOTTLENECKS
In addition to needing improved VOE manageability, IT at SMB sites is often constrained from evolving into the next stage of virtualization by the need to deliver
09
VDI Test Scenario
XIOTECHISE VM VIEW
Within Xiotech’s ISE Manager, Virtual View provides administrators with the means to globally configure storage volumes on VMs residing on hosts provisioned with storage volumes that were imported from an ISE. Having provisioned an 800BG volume (F: on Dell1900VMb) specifically for use by oblVD-VMM, our VM running System Center Virtual Machine Manager, we were able to configure a volume on that VM that resided on that specific host volume.
sufficiently high levels of application performance with respect to CPU, memory, and storage resources within their VOE. When resource constraints are storage centric, critical business continuity applications, such as backup and replication, are negatively impacted as throughput bottlenecks extend backup windows for multiple VMs consolidated on a host server.
With multiple VMs consolidated on a limited number of hosts, storage throughput bottlenecks cascade over multiple VMs, which puts a high premium on storage devices with low I/O latency. This is a particular strong point for the Emprise 5000 and its ISE technology, which presents a proprietary DataPac as a virtual super disk.
Using two virtual SCSI disks attached to a VM running Windows 7,
openBench Labs ran Iometer to stream full-duplex large-block (128KB) reads and writes. Each of the virtual disks was backed by a distinct volume on the host server. Throughput for both the read and write I/O streams averaged 475MB per second. As a result, combined full-duplex throughput maintained a steady-state level of 950MB per second. This result is double the level of sequential throughput that is required to support such specialized workstation and server applications as HD video editing, multi-dimensional On Line Analytical Processing (OLAP), data mining, and disaster recovery operations.
On the other hand, small-block (8KB) random access I/O is especially important for database-driven server applications. To set the bar for VM support of transaction processing (TP) applications built on databases such as SQL Server and Oracle, we used Iometer with the same VM and virtual SCSI disks.
VDI Test Scenario
HYPER-V VM SEQUENTIALI/O
To assess potential I/O performance of server and desktop VMs in our Hyper-V
environment, we ran Iometer on a VM hosted on the Hyper-V host dubbed Dell1900VMb. In this test, we simultaneously launched a stream of 128KB read requests and a stream of 128KB write requests on separate virtual SCSI disks. Each test disk on the desktop VM was backed by a separate volume imported from the Emprise 5000 by the host. Read and write requests were equally balanced over both FC ports on the host server and the Emprise 5000. The bottom line for performance was an I/O throughput level of 950MB per second for a single VM.
In our TP benchmark tests, we started with a single I/O stream that mixed 80 percent read requests with 20 percent write requests. As we observed during our large-block sequential access test, small-block TP requests were equally balanced across all FC ports on our SAN fabric. More importantly, on this test, we sustained 4,000 IOPS with an average access time of less than 12ms. We were then able to scale our benchmark nearly linearly with two streams to separate virtual drives on our VM, as we maintained an average access time of less than 20 ms while sustaining over 7,000 IOPS.
DESKTOPCONNECTIONS
The I/O throughput performance of Hyper-V-VMs is a significant plus for IT when setting up a VDI environment. The
sporadic nature of data access patterns generated by desktop applications is central to the problem of setting up storage for a VDI. Unlike server applications, which are often tuned for specific data handling, such as I/O streaming or random TP processing, desktop applications seldom optimize for I/O and rely on standard sequential file access using 8KB data blocks. As a result, I/O patterns for desktop applications produce minimal I/O stress and generate I/O loads that are well within the boundaries of most storage systems.
Within a virtual desktop infrastructure, however, the traditional desktop I/O utilization pattern scales in a way that leaves desktop VMs prone to severe degradation during unforeseen I/O storms. While I/O storms are sporadic events, they are statistically predictable. In particular, these events occur when multiple users simultaneously
implement administrative processes such as logging in, running a virus scan, or updating VDI Test Scenario
11
WINDOWS7 VM BUSINESSDESKTOP
We tracked I/O demands of desktop VMs while running a number of typical business applications. In all cases, work on desktop VMs, such as obl-VD1, put little stress on our Emprise 5000. Even while opening and rebuilding a large MS Access database, throughput spiked at 50MB per second and averaged 15MB per second.
While video performance was extraordinary with the ICA protocol, more important for business processing was the automatic setup of all drives on the initiating PC as mapped network drives on the VM desktop.
software on line.
Meeting the irregular nature of I/O patterns in a VDI demands that IT plan for VDI storage that can scale performance with respect to peak I/O activity, rather than typical user activity. In particular, as IT administrators densely provision multiple desktops VMs on a VOE host, the aggregate I/O load takes on the characteristics of random 8KB I/O requests. As a result, IT decision makers need to put a premium on highly-adaptive, low-latency, storage systems, such as the Emprise 5000 with its ISE technology, that can respond to sudden disruptive changes in I/O processing.
Customer Value
VIRTUALMACHINEFARMS
The key to garnering a positive ROI on any VDI initiative rests in simplifying the management and optimizing the utilization of storage, CPU, and memory, and resources. By far, the greatest payback for any VDI initiative will come out of improvements in the management and utilization of storage resources.
Within a Citrix Virtual Desktop Infrastructure, the central component is the Citrix Desktop Delivery Controller (DDC), which brokers user requests for desktop access. The key mechanism in this scheme is that of a farm, which consists of administrators, systems that are running the DDC, and defined groups of desktop VMs, which can be managed as single object.
When an IT administrator creates a desktop VM group, the desktop VMs are associated with end users or end-user groups via Active Directory. In this assignment process, an administrator can explicitly assign a desktop VM to an end user; allow the automatic creation of a permanent link to an end user on first use of the VM; or configure the Desktop Delivery Controller to randomly allocate a VM to an end user at each logon. Within the context of this role as a session broker, the Desktop Delivery Controller also starts and stops desktop VMs based on user demand and configuration rules created by IT administrators.
13
VDI Automation & ROI
VDI Automation & ROI
“A
key part of the Citrix Provisioning Services value proposition is
the utilization of SQL Server to maintain and stream a disk image,
which is dubbed a vDisk and used by all of the members of a Desktop
Group associated with a Desktop Delivery Controller as a boot disk.”
XENDESKTOP CONNECTIONBROKER
For our tests we created two desktop VM groups. The first group, dubbed MyPCpool, contained conventional VMs. Each of these VMs booted from a virtual disk (vDisk) that was created and configured for exclusive use by that VM. VMs in the second group, vdiVDGroup, all booted over the network using a shared vDisk image streamed from the server running Citrix Provisioning Services.
14
End users request desktop services via the Citrix Online or Web Plugin utilities, which are installed on any system that will function as a client. The end user provides the Internet address of a server running DDC and Active Directory credentials.
Alternatively, the Citrix Plugin can be configured to automatically pass the user’s credentials established on the client device. Based on the login information provided by the Citrix Plugin, DDC returns a list of Desktop Groups, which are pools of VMs that are available for that user. In turn the Citrix Plugin displays the list of available Desktop Groups as part of its user interface.
When the end user chooses a Desktop Group, the Desktop Delivery Controller brokers a connection with the host server for an appropriate desktop VM. The host server starts the VM and ICA communications are established between the Citrix Plugin and the Virtual Desktop Agent that is running on the VM.
STREAMINGAUTOMATION
The Citrix VDI architectures also provides a mechanism to automatically create and provision a group of virtual desktops using a single common disk image that is used to boot each desktop VM in the designated group. This method of provisioning optimizes storage utilization, simplifies OS change and security management tasks, and provides end users with a clean virtual desktop each time they log on.
Automating Citrix Provisioning Services (PvS) in a Hyper-V environment requires running MS System Center Virtual Machine Manager (SCVMM). In addition, to enable Citrix DDC and PvS to use Hyper-V and SCVMM, the SCVMM management console must be installed on a sever before installing either Citrix module.
In our test case, we created a single 75GB vDisk from which we booted 10 diskless VMs saving 350GB of storage. By using a dynamic rather than fixed vDisk, we further reduced the size of the boot image to just 18GB and increased storage capacity savings to 570GB.
Nonetheless, the big contribution to ROI, however, does not come from the substantial space savings and storage optimization. By using a common boot disk, we could update
VDI Automation & ROI
AUTOMATEDDESKTOPPROVISIONING
After creating a copy of a VM system disk in the PvSStore, we were able to assign that image to a PvS device collection, vdiVDGroup, that automatically configured VMs to boot from that image and placed those VMs into a corresponding DCC desktop group.
and refresh the operating system or any applications installed on all of the VMs in the vdiDTGroup desktop group on the DCC updating the VM that served as the template and then updating the image of that VM’s boot disk.
More importantly, using the diskless VM model with Citrix Provisioning Services significantly changes I/O throughput characteristics. All data must now pass through SQL Server with all communications between the desktop VM and SQL Server moving over the LAN. As more VMs utilize vDisks as their boot devices, I/O access and throughput for SQL Server data stores
maintaining these devices become critical issues for maintaining site operations.
What’s more, SQL Server constantly updates internal tables with on-going operations-oriented metadata. This makes it essential for IT to provide these database servers with fast highly reliable storage in order to support a quick recovery time should a database become corrupted. In particular we were able to
leverage the ability of the Emprise 5000 to support full-duplex streaming of data and run backups of the PvS database of boot images at upwards of 180MB per second.
VDI Automation & ROI
15
PVS DATABASEBACKUP
On the Hyper-V host, we leveraged the throuput capabilities of the Emprise 5000 to back up the full PvSStore at upwards of 190MB per second.
BETTERBUILDINGBLOCKS
Xiotech’s ISE technology provides an innovative solution for improving the reliability and performance of disk-based storage systems. By approaching disk drives as a grid of storage surfaces, Xiotech has enabled the controllers in the Emprise 5000 to go beyond simply accessing data and actively manage component reliability.
More importantly, Xiotech’s change to the underlying technology of storage systems transforms the notion of a basic storage building block for both IT and OEM users in a way that makes storage provisioning synergistic with the Service Level Agreements (SLAs) set up between IT and Line of Business divisions. The sophisticated characteristics of ISE technology rather than simple electronic specifications, define building blocks that are application-centric rather than connection-centric.
The Emprise 5000 storage system builds on ISE technology to eliminate the need for maintenance intervention by IT administrators and to provide near-linear scaling of application throughput metrics as the number of storage systems
increases. What’s more, ISE Manager software provides the means to manage multiple Emprise 5000 devices as a single logical cloud. As a result, IT administrators are able to deploy multiple Emprise 5000 systems, to cost-effectively meet and support Service Level Agreements for multiple application-centric environments, including Virtual Operating Environments for both servers and desktop systems.
While reducing OpEx and CapEx costs are the critical divers in justifying the acquisition of storage resources, those resources must first and foremost meet the performance metrics needed by end-user organizations and frequently codified in SLAs. In terms of common SLA metrics, our benchmarks for a single Emprise 5000 reached levels of performance that should easily meet the requirements of most applications.
Customer Value
Customer Value
“H
ost-based ISE Manager software enables multiple Emprise 5000
units to be managed as a single ISE cloud, which includes
provisioning VMs on VMware, Hyper-V, or Citrix Xen host servers.”
16
X
IOTECH
E
MPRISE
5000 F
EATURE
B
ENEFITS
1) Application-centric Storage: DataPacs tuned to application needs define the base storage unit rather than individual disk drives. 2) Minimize Storage Management Costs:Host-based ISE Manager
software enables multiple Emprise 5000 units to be managed as a single ISE cloud, which includes provisioning VMs on VMware, Hyper-V, or Citrix Xen host servers.
2) Autonomic Self-healing Storage System:Integrated firmware on drives and controllers allows fine-grain data access to be defined at the level of drive heads in order to boost I/O throughput and automatically repair and rebuild drives in-situ.
3) High Streaming Throughput:Running with Balanced DataPacs, openBench Labs benchmarked streaming read performance at 380MB per second and streaming write performance at 175MB per second on VMs running Windows 7.
4) Linear Scaling of IOPS: Using random 8KB I/O requests in an 80/20 mix of reads and writes sustained 3500 IOPS with an average response time of less than 25ms. Using two VMs on separate DataPacs, performance scaled to 6,500 IOPS.
