DOCX

4AA1_4257ENW_Veritest_Feature_and_Functionality

By Elsie Anderson,2014-04-23 15:13
14 views 0
4AA1_4257ENW_Veritest_Feature_and_Functionality

    May 2010

    HP StorageWorks P4500 SAS SAN vs. Dell EqualLogic PS6000XV:

    Feature and Functionality Comparison Test report prepared under contract from Hewlett-Packard Executive summary

Hewlett-Packard (HP)

    commissioned VeriTest, the testing

     service of Lionbridge Technologies,

    Inc., to compare features and The 3-node HP P4500 SAN solution provides up to 32% functionality between the HP greater usable storage capacity, (see pg 12). StorageWorks P4500 SAS SAN

     The 3-node HP P4500 SAN solution provides better High Solutions and the Dell PS6000XV

    Availability features than the 2-node Dell PS6000XV SAN storage arrays. The 3-node HP

    solution - at a 14% lower price. (see pg 1) P4500 SAN, a Virtualization SAN

    with an HP P4500 expansion node, The 3-node HP P4500 SAN solution provides greater has a current list price of $96,600.00. availability than the Dell EqualLogic 2-node PS6000XV The 2-node Dell PS6000XV currently solution. Under the three failure scenarios tested, NIC, Node has a list price of $110,000. The goal and Double-disk, the HP P4500 SAN experienced no of this study was to evaluate these downtime, whereas the Dell PS6000XV experienced as much devices in terms of high availability, as 30 minutes of downtime (see pgs 8-11). capacity utilization under various

    common usage scenarios, and ease When retaining data via snapshots the 3-node HP P4500

    of upgrading the storage solution to SAN solution has as much as 62% lower storage overhead,

    meet future storage requirements. than the Dell EqualLogic 2-node PS6000XV (see pg 15).

     With its High Availability features, the 3-node HP P4500 SAN Both HP and Dell provide a scalable complements the VMware vSphere 4 Fault Tolerance iSCSI based solution, each claiming feature, whereas the Dell PS6000XV does not provide High the following benefits: scalability, Availability (see pg 13) flexibility, cost effectiveness and high

    availability. Lionbridge developed a Platform refresh of a Dell PS6000XV SAN take 10 times as

    test methodology that compared long as for a HP P45000 SAN 5 hours 39 minutes vs. 34

    these benefits between the two minutes (see pg 18).

    vendors to help end users better understand the real-world differences

    experienced with each product under

    test. Both solutions are based on a scale-out technology, combining multiple nodes into a cluster and thus

    presenting a common storage pool. For these comparable prices, the configurations below are what were available for each product. Therefore, the hardware configurations were not identical with regard to number

    of nodes and disk spindles.

The HP P4500 SAN configuration consisted of a three (3) node cluster running SAN/iQ version 8.1.00.0047.0

    with RAID5, and 2-way replication. Each node contained 12 450GB 15K RPM SAS drives with a total raw capacity of 16.2TB.

    The Dell PS6000XV configuration consisted of a two (2) node cluster running firmware version 4.1.3 with RAID5 striped across both nodes. Each node contained 16 450GB 15K RPM SAS drives, with dual controllers, and total raw capacity of 14.4TB.

    To identify the differences between the two products under test, Lionbridge developed a test methodology that executed several common scenarios encountered by storage administrators. To best highlight these scenarios, Lionbridge used a fictitious company, called ACME Corp, and identified several requirements for comparing and contrasting the differences between the two products under test.

    ACME Corp is a hypothetical mid-sized company providing storage for online records management. They are looking to upgrade their existing SAN solution to an iSCSI based centralized storage solution that will grow with the company needs and integrate with their existing VMware ESX environment. Aside from meeting current storage capacity requirements and future storage growth needs, High Availability (“HA”) is of most

    importance to ACME Corp. ACME Corp has Service Level Agreements (SLA) with their customers in which any outage (defined as the customer being unable to access the data) results in penalties ACME Corp pays to their customer(s) affected by the outage. For ACME Corp every minute of downtime results in a $1,000 penalty paid to each customer affected by the outage.

    For each type of enterprise, there is a specific cost to any service interruption, of any length. But the general question "What does an outage cost?" has no meaningful answer. In this research, Gartner develops a method for estimating the cost of an outage for a particular organization. Read the full report "Q&A: How

    Much Does an Hour of Downtime Cost?" from 9/29/2009, to find out more.

    In addition to HA, ACME Corp employs snapshots as a means to maintain readily available online backups to transactions. Therefore, the ability to maintain several snapshots, as well as the ease with which the process is managed are key factors in determining the storage solution selected. For Disaster Recovery (“DR”) purposes, ACME Corp has a sister site located several miles away to act as a storage replication target. Remote replication efficiency is another item of importance to ACME Corp for determining the storage solution selected. Finally, as with many other companies, ACME Corp is facing the rapid growth in storage space requirements. Due to this fact, it is also important that ACME be able to easily and rapidly upgrade or scale the storage solution to meet additional capacity requirements.

    Although the focus of this white paper’s comparison is not on platform performance, performance is an important consideration when selecting the storage solution. To help end-users determine their options for backend storage when deploying an Exchange server, Microsoft maintains a benchmark for Exchange called ESRP that is a good proxy for random IO. Both vendors have published ESRP results, click here for further

    details or contact the vendors directly to discuss performance for other types of workloads. VMware also hosts a website with performance information, click here for further details.

    To assist the end-user, Table 1 below highlights system specifications associated with platform performance capabilities.

     HP StorageWorks P4500 SAN vs. Dell EqualLogic PS6000xv Feature Comparison 2

    Item HP P4500 SAS Dell PS6000XV

    Nodes 3 2

    Controllers 3 Active / 0 Passive 2 Active / 2 Passive

    CPUs 3 Active / 0 Passive 2 Active / 2 Passive

    RAID Configuration 6 RAID-5 (5+1) raid sets 4 RAID-5 + Hot Spare (6+1+1)

    raid sets

    Disks 36 15K 450GB SAS disks 32 15K 450GB SAS disks

    36 of 36 disks active for 28 of 32 disks active for

    performance performance

    Formatted Usable Capacity 12.03 TB 10.98 TB

    Cache / RAM 7.5 GB Active / 0 GB Passive 4 GB Active / 4 GB Passive

    Bandwidth 6gbit (6 x 1gbit active) 8gbit (8 x 1gbit active)

    Volume Connectivity iSCSI Standard iSCSI Standard

    Volumes Supported / group 1500 1024

    Hosts Supported / group 4000 512

    Max Disks Supported / group 384 256

    Table 1: Performance Related Platform Specifications

    Testing revealed that the 3-node HP P4500 SAN Solution provided greater usable storage capacity, while also providing better high availability features than the 3-node Dell PS6000XV solution at a lower price point.

    Testing also revealed that the 3-node HP P4500 SAN solution provided lower storage overhead when retaining data via snapshots and remote replication than the 2-node Dell PS6000XV solution.

    The 3- node HP P4500 SAN solution required fewer steps and only 1/10 the time required to perform a platform refresh as compared to the 2-node Dell PS6000XV solution.

    Finally, the high availability features of the HP P4500 SAN complemented the VMware vSphere 4 Fault Tolerance feature, thereby providing a better HA solution for ACME Corp.

Testing methodology

     ACME Corp is looking to upgrade their existing SAN solution to an iSCSI based storage solution. They need a solution that meets the following criteria:

    ; Minimum usable capacity of 10 TB

    ; Provide the ability to survive the most common of hardware failures (i.e. network, system, controller

    and disk failures)

    ; Provide snapshot capabilities for online point-in-time data retention

    ; Provide remote replication capabilities for disaster recovery requirements

    ; Be compatible with and integrate with VMware vSphere 4, and specifically work with the Fault

    Tolerance configuration capabilities

    ; Provide an ease of upgrade path for future-proof needs

    With the above requirements in mind, the HP P4500 SAN solution and the Dell PS6000XV arrays were viable options to consider.

    Lionbridge configured a test environment that consisted of eight (8) HP BL460c G6 Blade Servers in a c3000 enclosure. The c3000 enclosure (backplane is shared) included the following:

    ; Two (2) Flex-10 interconnects shared by all blades

    ; One (1) 10GigE uplink for the internet connection

    ; One (1) 10GigE uplink for dedicated iSCSI traffic

     HP StorageWorks P4500 SAN vs. Dell EqualLogic PS6000xv Feature Comparison 3

    For network connectivity a single 48 port HP ProCurve 3400cl was used. This switch contained two VLANs, one that was used for administrative, remote desktop protocol (RDP) access and internet connectivity. The second VLAN was configured to provide dedicated iSCSI data traffic.

    One of the blades was configured with Microsoft Windows Server 2008 R2 x64. This blade was used to manage the storage arrays under test. It had the HP Centralized Management Console (CMC) application

    installed to manage the HP P4500 SAN solution and the Dell EqualLogic Group Manager GUI to manage the Dell EqualLogic arrays.

    Two blades were configured with VMware vSphere 4 as ESX host servers. Each ESX host was configured with a single Windows Server 2008 R2 x64 Virtual Machine (VM). Each VM was configured with two (2)

    vCPUs and 4 GB RAM. A single NIC was dedicated for virtual machine traffic on each host. One ESX host/VM was configured to connect to the P4500 SAN solution, the other configured to connect to the PS6000XV array. The ESX configuration was modified to connect to their respective storage solution using the built-in ESX software iSCSI initiator. IQN authentication was used to manage the iSCSI connections. The VM image was contained on a 250 GB volume, thick provisioned on each storage solution for their respective VM.

    The HP P4500 SAN was configured using three (3) nodes. The cluster was configured into disk RAID5 with Network RAID configured in 2-way replication (as of publication known as Network RAID 10) for all volumes. Each P4500 SAN node was configured with a single 802.3ad bond using both embedded NICs.

    The Dell PS6000XV was configured with a two (2) node, RAID5 striped configuration. No synchronous data replication was configured as this was not a feature available at the time of testing. Each EqualLogic PS6000XV node was configured with its default active/passive controller configuration, and each controller had four (4)1GigE network ports configured.

    For each of the test cases listed below, a single 100GB volume was created on each storage solution and presented to their respective VMs as a second hard disk drive via the ESX software iSCSI initiator. This feature was enabled in the ESX configuration and was setup to use IQN authentication. For the VMware Fault Tolerance test case only, the storage arrays were configured to permit multiple client connections to allow two ESX hosts to connect to the provisioned storage.

    The industry standard benchmark utility IOMeter was used to generate workload to the 100GB LUN for each test case. The LUNs were kept in a raw, non-formatted state. All of the tests used an OLTP-like workload that online database applications typically encounter. Like many other industry standard OLTP benchmarks, including the majority of the top 10 TPC-C price/performance results, we used an 8KB request size with a mix of random read and write operations. IOMeter was configured with 8 outstanding IOs, 8KB request size, a mixture of 60% random read and 40% random write, with an 8KB alignment. The IOMeter workload was started five (5) minutes prior to each test case, and was allowed to run through completion of the test case.

Figure 1 below provides an overview of the test environment.

     HP StorageWorks P4500 SAN vs. Dell EqualLogic PS6000xv Feature Comparison 4

     Figure 1: Test Environment Network Topology

    Several test cases were developed to highlight key characteristics and to compare the HP P4500 SAN solution with the Dell PS6000XV storage arrays. These test cases were designed to address the requirements listed above for the fictitious ACME Corp. Each test case methodology is detailed in the sections below.

    High Availability Comparison

    It is important to understand what effect a hardware based failure can have on data availability. There

    are numerous points in the architecture of any storage solution where a hardware failure can occur. This

    section covers several common hardware failure scenarios. Lionbridge Test Engineers executed these

    hardware failure tests, and observed for the effect the failure had on the overall availability of the data.

    This section is broken into several subsections that focus on the specific hardware failure simulated.

    NIC Failure

    This test case was designed to observe behavior of each storage solution when a network failure

    occurred. In this case, a network failure was introduced by removing all network cables to one of the

    storage nodes. Hardware-based NIC only failure (for multiple NICs) is a very rare event, likely resulting

    from human error. Under a normal Best Practice configuration, the NICs would be connected to multiple

    switches which reduce the risk of a hardware failure causing the outage. For purposes of testing, this

    “human error” event was compared to see how each storage solution behaved under such an incident.

     HP StorageWorks P4500 SAN vs. Dell EqualLogic PS6000xv Feature Comparison 5

    For this test case, a test engineer pulled all four cables from the primary controller on one of the EqualLogic nodes at the same time. Next, the network cables were pulled from one of the HP P4500 nodes to simulate the same type of failure on that platform. The test engineer observed the behavior of the VM and the IOMeter application. Also observed was the “Group Monitor” on the EqualLogic Group

    Admin GUI, as well as the “Performance Monitor” on the CMC. Each management application was also

    monitored for any errors reported. After five (5) minutes the network cables were reconnected and the arrays were allowed to return to optimal operation.

    The ESX host for each VM was also monitored using vCenter. In addition, an active RDP connection to the VM was used to verify VM availability and IOMeter workload continued to function. System Failure

    This test case was designed to observe the behavior of each storage solution when a single node failure event occurred. Such a failure event can occur due to a number of reasons, such as power supply failures, system board failure, human error, etc.

    For this test case, a test engineer simulated a system failure by pulling the power cable(s) from one of the nodes in each platform. The test engineer observed the behavior of the VM and the IOMeter application. Also observed was the “Group Monitor” on the EqualLogic Group Admin GUI, as well as the

    “Performance Monitor” on the CMC. Each management application was also monitored for any errors

    reported. After five (5) minutes the power cables were reconnected and the arrays were allowed to return to optimal operation.

    The ESX host for each VM was also monitored using vCenter. In addition, an active RDP connection to the VM was used to verify VM availability and IOMeter workload continued to function. Double-disk Failure

    This test case was designed to observe the behavior of each storage solution in the event a double-disk failure (two disks on the same node) event occurred. Although relatively rare, double-disk failures do occur and it is important to understand the impact such an incident has on the availability of data.

    For this test case, a test engineer simulated a double-disk failure by pulling two disks from the same node at the same time. Two odd numbered disks were pulled from one of the P4500 SAN nodes, as well as two odd numbered disks from one of the EqualLogic PS6000XV nodes. The test engineer observed the behavior of the VM and the IOMeter application. Also observed was the “Group Monitor” on the

    EqualLogic Group Admin GUI, as well as the “Performance Monitor” on the CMC. Each management application was also monitored for any errors reported. After five (5) minutes the disks were re-inserted and the arrays were allowed to return to optimal operation.

    The ESX host for each VM was also monitored using vCenter. In addition, an active RDP connection to the VM was used to verify VM availability and IOMeter workload continued to function. VMware Fault Tolerance Scenario

    This test case was designed to demonstrate the HA capability of the storage solution when running within a VMware vSphere 4 ESX cluster with Fault Tolerance (FT). A two-node ESX cluster was configured

    with Fault Tolerance enabled. This cluster was configured to have access to both the P4500 SAN and PS6000XV storage clusters.

    Future plans for ACME Corp are to move to a multi-site configuration in order to better provide data availability. By configuring into a multi-site configuration, the HA capabilities are more fully leveraged so that single points of failure (power distribution units, network infrastructure, racks, buildings, etc) are mitigated even further than the current configuration.

    For this test case, a test engineer simulated a system failure and an ESX host failure at the same time. To accomplish this, the engineer removed the power cables from one of the ESX hosts as well as one of the nodes from each of the storage arrays. Testing was conducted serially. Power was pulled from the

     HP StorageWorks P4500 SAN vs. Dell EqualLogic PS6000xv Feature Comparison 6

    ESX host first to allow the Fault Tolerance failover to occur. As this was running, the power was then pulled from one of the storage nodes under test. VMware FT was configured so that one ESX host was running the primary VM for the HP P4500 SAN and the other ESX host was running the primary VM for the Dell PS6000XV SAN. Prior to initiating the failure, the IOMeter OLTP script was executed on the VM to generate workload at the time of the failure.

    To verify that the VMware FT configuration was sound, a baseline test was executed for each storage solution in which just one of the ESX hosts was failed by removing power to it. This initiated the FT failover process, and the Lionbridge Test Engineer verified the VM failed over properly and no outage was experienced by the running VM or the IOMeter OLTP workload that was generating active I/O at the time.

First, the P4500 SAN was tested by removing the power from the ESX host running the “live” (or primary)

    VM. Once failure was detected and the test engineer confirmed failover was initiated, the power was removed from one of the P4500 SAN nodes in the cluster. The test engineer then observed VMware vCenter and the VM console as well as the CMC GUI for any errors.

    Second, the EqualLogic PS6000XV was tested by removing the power from the ESX host running the “live” (or primary) VM. Once failure was detected and the test engineer confirmed failover was initiated, the power was removed from one of the PS6000XV nodes in the cluster. The test engineer then observed VMware vCenter and the VM console as well as the Group Admin GUI for any errors.

     Overall Capacity Utilization

    ACME Corp has a current storage capacity requirement of 10TB. This is to cover existing data storage needs as well as expected growth over the next 12-14 months. Due to common budget approaches, the next opportunity to add additional storage will come during the next fiscal year.

    This test case reviewed the available (usable) storage under different RAID configuration options available for each storage platform under test. This was used to determine the RAID options available to meet the minimum capacity requirements for ACME Corp of 10TB.

    For this test case, the test engineer simply configured each storage array in the various RAID types offered in the platform to determine the available capacity for each type.

     Capacity Utilization with Snapshots

    This test case compared the amount of storage provisioned by each platform for snapshot retention. Also compared was the process of taking snapshots and the observed results.

    For this test case, the test engineer created a 100GB volume, presented it to the respective VM as a second HDD as a raw device, and then filled it to 100% capacity using the IOMeter benchmark utility. For this data fill operation, a sequential write workload was configured for IOMeter with the following settings:

    ; Outstanding IOs: 8

    ; 100% Sequential Write

    ; 64K Request size

    ; 64K Alignment

    An initial snapshot was then taken of the base data. IOMeter was run again using the OLTP workload to cause approximately 5% in data change. Then another snapshot was taken. This process was repeated a second and third time. For each of the iterations executed, the amount of storage provisioned to retain the snapshots was compared.

    The EqualLogic solution was compared with both 100% snapshot reservation and also with a 200% reservation. This was required due to the manner in which the Dell EqualLogic handles snapshots by default. When snapshot reservation is exhausted, the default action with the PS6000XV is to remove the oldest snapshot. The Lionbridge Test Engineer observed that with 100% reservation, which is the default for the PS6000XV series, not all snapshot iterations could be retained following the above methodology.

     HP StorageWorks P4500 SAN vs. Dell EqualLogic PS6000xv Feature Comparison 7

    Therefore, it was necessary to run a second test for the PS6000XV solution with a higher reservation

    amount so that all snapshots could be retained. This auto-deletion setting led to a situation in which a

    snapshot that was intended to be retained, was auto-deleted. It is important for an end user to

    understand the repercussions of this setting in order to make sure that required snapshots are not deleted

    inadvertently by the system.

     Remote Replication Efficiency

    This test case compared the amount of storage provisioned by each platform for remote replication

    retention. Also compared was the process of configuring and initiating remote replication.

    For this test case, the test engineer created a 100GB volume, presented it to the respective VM as a

    second HDD as a raw device, and then filled it to 100% capacity using the IOMeter benchmark utility. For

    this data fill operation, a sequential write workload was configured for IOMeter with the following settings:

    ; Outstanding IOs: 8

    ; 100% Sequential Write

    ; 64K Request size

    ; 64K Alignment

    The solution was configured with a single node, RAID5, no replication. A second cluster was created,

    containing a single node, also configured as RAID5, no replication. An initial remote copy operation was

    executed after the volume had been fully written. Then a 5% data change was introduced using the

    IOMeter OLTP workload. This was repeated a second and third time, each time a new remote copy was

    taken. The total amount of provisioned storage on the remote node was recorded.

    The EqualLogic solution was configured with a single (source) node group in RAID5. A second (target)

    group was configured with a single node in RAID5, and the replication partnership was configured for

    one-way replication from the source to the target group.

    A thin provisioned 100GB volume that was filled to 50% capacity was created. Then, an IOMeter OLTP

    workload was executed to cause 5% change in data. These changes were replicated to the replication

    target for each platform and the amount of provisioned space was compared.

    Platform Refresh

    This test case compared the process of refreshing the existing storage solutions to newer product

    offerings. As time goes by, storage capacity must increase to meet ever changing business requirements.

    It is imperative that any storage solution provide a means to update the underlying hardware without

    taking data offline. The process to accomplish the node upgrade and the amount of time required was

    compared.

    For this test case, the P4500 cluster was configured with a two-node solution, RAID5, and no replication.

    This was then “upgraded” by adding two new nodes of equal capacity to the cluster and removing the “old”

    nodes.

    The EqualLogic cluster was configured with a single node with RAID5. A second node was added to the

    group and then after the raid verification process completed, the “old” node was vacated.

    Test results

    This section provides all the results for the high availability, capacity utilization and feature comparison along with usability as experienced by the Lionbridge Test Engineer using the HP P4500 SAN and Dell PS6000XV storage arrays. Please refer to the testing methodology section of this report for complete details on how we conducted these tests.

     High Availability Comparison

    This section covers the results for test cases simulating several common hardware failure scenarios.

    Lionbridge Test Engineers executed these hardware failure tests and observed the effect the failure had

     HP StorageWorks P4500 SAN vs. Dell EqualLogic PS6000xv Feature Comparison 8

    on the overall availability of the data. This section is broken into several subsections that focus on the specific hardware failure simulated.

    It is important to note that the HP P4500 solution provides the same level of High Availability in a 2-node solution as it does with the 3-node solution using 2-way data replication. However, due to the less expensive acquisition cost, we were able to add a third node to the configuration that results in greater usable capacity, at a lower cost than the Dell EqualLogic 2-node solution. By adding a third node to the configuration, the HP solution meets all of the HA, capacity and feature requirements identified for ACME Corp.

    High Availability Test Case Results

    Table 2 below provides a summary of the results for each of the HA tests conducted by Lionbridge. Please refer to each test section below for detailed results.

     HP P4500 SAS Dell PS6000XV

    HA Test Result Downtime Cost Result Downtime Cost

    NIC Failure Volumes Online $0 Volumes Offline $10K

    System Failure Volumes Online $0 Volumes Offline $13K

    Double-disk Failure Volumes Online $0 Volumes Offline $30K

    Table 2: High Availability Test Case Results - Summary

    NIC Failure

    For this test case, a test engineer pulled all four cables at the same time from the primary controller on one of the EqualLogic arrays. Next, both network cables were pulled from one of the HP P4500 nodes to simulate the same type of failure on that platform. The test engineer observed the behavior of the VM and the IOMeter application. Also observed was the “Group Monitor” on the EqualLogic Group Admin

    GUI, as well as the “Performance Monitor” on the CMC. Each management application was monitored for any errors reported. After five (5) minutes the network cables were reconnected and the arrays were allowed to return to optimal operation.

    The Lionbridge Test Engineer observed the following behavior for the HP P4500 SAN solution:

    ; The CMC GUI reported the effected node as offline

    ; 2-way replication went into a “degraded” state

    ; VMware ESX host did not report any issues

    ; The Virtual Machine continued to run without issue

    ; The IOMeter session running in the VM continued to execute without error

    The end result of this test case was that under a NIC failure scenario on the P4500 SAN no outage was experienced. Data remained online and accessible during the entire time the node was in the failed state.

    The Lionbridge Test Engineer observed the following behavior for the Dell PS6000XV solution:

    ; The Group Admin GUI reported the affected node as offline

    ; All striped volumes went offline

    ; The VMware ESX host reported connectivity problems to the VM images and IOMeter DataStors

    ; The Virtual Machine hung, and was unresponsive

    ; The Group Admin “Group Monitor” did not show any activity, and iSCSI sessions from the ESX

    host were disconnected.

    The end result of this test case was that under a NIC failure scenario on the EqualLogic PS6000XV the data was offline and an outage had occurred.

    After the cables were re-connected to the NICs, the Group Admin GUI reported the volumes back online within several seconds of the reconnect. After the cables were re-connected to the NICs, an HBA rescan on the ESX host had to be run in order to re-establish iSCSI connectivity and to bring the VM back online.

     HP StorageWorks P4500 SAN vs. Dell EqualLogic PS6000xv Feature Comparison 9

    This total process led to a “downtime” of approximately 10 minutes, including the 5 minutes the cables were left unplugged.

    Experiencing such a downtime can lead to significant SLA penalties for ACME Corp. The downtime experienced above would equate to approximately $10,000 in SLA penalties with the Dell PS6000XV, but none with the HP P4500.

    Figure 2 below shows the downtime cost associated with the above example when a NIC failure occurred.

    SLA Penalty -NIC Failure

    $10,000

    $10,000

    $8,000

    $6,000

    $4,000

    $0 $2,000

    $0

    HP P4500Dell PS6000XV

     Figure 2: NIC Failure Outage Cost in SLA Penalties

    System Failure

    For this test case, a test engineer simulated a system failure by pulling the power cable(s) from one of the nodes from each vendor. The test engineer observed the behavior of the VM and the IOMeter application. Also observed was the “Group Monitor” on the EqualLogic Group Admin GUI, as well as the

    “Performance Monitor” on the CMC. Each management application was monitored for any errors reported. After five (5) minutes the power cables were reconnected and the arrays were allowed to return to optimal operation.

    The Lionbridge Test Engineer observed the following behavior for the HP P4500 SAN solution:

    ; The CMC GUI reported the affected node as offline

    ; 2-way replication went into a “degraded” state

    ; The VMware ESX host did not report any issues

    ; The Virtual Machine continued to run without issue

    ; The IOMeter session running in the VM continued to execute without error

    The end result of this test case was that under a single node failure on the P4500 SAN, no outage occurred. The data remained online and accessible during the entire time the node was in the failed state.

    The Lionbridge Test Engineer observed the following behavior for the Dell PS6000XV solution:

    ; The Group Admin GUI connection was lost and an automatic re-connection was initiated. It took

    approximately 70 seconds for the reconnection to complete.

    ; The node that was failed showed as offline in the Group Admin GUI

    ; All striped volumes, which included the VM image and IOMeter data LUN, were listed as offline

    ; I/O had stopped according to the Group Admin “Group Monitor” link

    ; The VM hung and new connections were unavailable.

     HP StorageWorks P4500 SAN vs. Dell EqualLogic PS6000xv Feature Comparison 10

Report this document

For any questions or suggestions please email
cust-service@docsford.com