MSA 1060/2060/2062 FAQ: System and Architecture Questions
MSA 1060/2060/2062 FAQ: System and Architecture Questions
Question: What are the 10GBase-T options for the MSA 1060/2060/2062?
Answer: 10GBase-T will be available on the MSA 1060 platform at initial NPI to replace 1GbE Copper connectivity.
Initially, the MSA 2060/2062 will use SFP+ optical/DAC iSCSI connections. 10GBase-T is being studied for
inclusion in MSA 2060/2062 product lines after the initial Gen6 launch.
Question: What are the SFP+ (Optical) connectivity options for the MSA 1060/2060/2062?
Answer: MSA 1060 iSCSI Systems will NOT offer a 10GbE SFP+ Optical connection. Customers wanting this
combination will need to move up to MSA 206x. The MSA 2060/2062 will use SFP+ optical/DAC iSCSI
connections.
Question: What cabling options are offered for the SFP+ (optical) connections on MSA 206X?
Answer: For the MSA 206x iSCSI systems, Optical SFPs and cables are supported along with DAC cables and new
QSFP breakout (1 to 4) cables are also supported. All supported cables are listed in the MSA 2060/2062
Quick Specs.
Question: Do MSA 2050 SAN controllers support FC and iSCSI at the same time?
Answer: No. For MSA 2050/2052 we supported iSCSI/FC on one Converged SAN controller. The I/O Chips used to
make this happen are no longer being manufactured by any vendor. Gen6 arrays, 1060/2060/2062 use
separate SKUs and controllers to offer iSCSI or FC or SAS.
Question: It is becoming mandatory on bids and tenders to have 32Gb Fiber Channel support on new arrays.
What should we bid?
Answer: On Gen6, we have the ability to support new protocols/link rates via a simple host interface module in
our controllers. A 32Gb FC module is not currently in process but the MSA team is investigating the
market opportunity for 32Gb. Adoption is very low currently due to switch and HBA costs. Until such a
time as MSA has a formal plan, bid Nimble.
Answer: No. 12 Gb SAS is made up of 4 x 12Gbs physical lanes. Fan-out cables take the four physical lanes of the
standard SAS cable and divide them by two. The connectivity to the host is 2 lanes @ 12Gb/sec. reaching
a Maximum sequential throughput of ~2.4Gb/sec per 2 wide cables.
Question: Is there any chance to have Y or fan-out cable on MSA206x?
Answer: Not planned. The GUI implementation is a challenge offering little benefit. 4 SAS ports/controller is a
natural upper-bound due to SAS HBA port counts and crossover cabling.
Question: MSA’s connect via SAS sometimes need to be connected to older servers. How can I tell which arrays
can be connected via which Smart Array controllers? Why are older Smart Array controllers not
certified on newer severs? And Vice versa?
Answer: The Compute Shared Options roadmap for supported SAS HBA and Smart Array controllers is driven by
that group. By policy, they don't test newer controllers to older servers nor older controllers on newer
servers to cut development and sustaining engineering costs on both platforms. The MSA Development
team adheres to their roadmap and lifecycle dates. Always refer to SPOCK for current support
information (https://h20272.www2.hpe.com/spock/index.aspx).
Question: How many GB’s controller cache are available on Gen6 controllers? How about ASICs?
Answer: 12Gb Cache per controller, 24GB Cache per system. Each controller contains a RAID acceleration ASIC.
Question: Any though or plan to add NVMe caching rather than using disk slots for Read Cache and Tiering?
Answer: MSA 6th Gen arrays do not have any internal integration for NVMe. Performance tier and Read-cache
are achieved by SAS SSD (RI) drives installed in the enclosure slots.
Question: Can Disk Enclosures be connected between MSA Gen5 and Gen6 systems?
Answer: This subject is covered nicely in training materials. Gen5 disk enclosures and all media are NOT
compatible with Gen6 arrays. Additionally, Gen6 Disk Enclosures and Media are NOT compatible with
Gen5 arrays.
Question: Are the MSA Gen6 HDDs compatible with Gen 10/Gen 10+ HDDs?
Answer: No, The Gen 10/Gen10+ HDDs are based on Smart Carriers. They will not fit in an MSA (Gen 1-6).
Question: Will MSA Gen6 drive SKUs have 6-Pack Drive Bundles again? Would it be possible to embed the ala
carte’ drive capacity in the Bundle’s SKU name?
Answer: Yes, 6-packs will be offered again. We will be driving more of a preference toward the Bundles moving
forward. At list price, 6-Packs will have a 25% delta to the single drives at a “per drive” level. We are
looking into the SKU naming suggestion.
Question: What are the addition software features coming with MSA Gen6?
V3 – September 2020
Answer: The Advanced Data Services SW Suite (ADS) is still the only SW license available on MSA Gen6. It
includes Performance Tiering, Snapshot increase for 64 (std) to 512 and Remote Snapshot Replication.
New for Gen6 are some enhancements to the Snapshot functionality which allow the system to now do
failover and failback. Additionally, the Performance Tiering engine gets a substantial revamp, termed
Tiering 2.0, which streamlines the page move algorithm to be more effective with incoming Writes and
with how it deals with free space on the SSDs. The results – The new Tiering 2.0 system delivers up to
45% more performance than the same workload run against a Gen5 array. The Tiering 1.0 or Tiering 2.0
engines are both very, very efficient and deliver amazing application acceleration to a wide variety of
workloads. Responding to dynamically changing workloads is this system’s forte. Though time, we have
integrated some tools into the tiering ecosystem which can help analyze hybrid system dynamics from
inside the GUI. Lear about the I/O Workload tool. While not a licensed feature, the new MSA-DP+ data
protection capability is very important going forward. It is covered in depth in our trainings materials,
whitepapers, etc. We will build on this new RAID type for years to come. Dig in and understand it.
Question: Does the new Gen6 MSA allow online drive FW upgrades w/o shutting down all hosts?
Answer: Not yet. It is on the development roadmap. This feature is now much easier to implement than in the
past when the HDD/SSD vendors had differing levels of support for online FW updates.
Question: Are there any plans to offer DeDup or Compression on the MSA platform?
Answer: Not currently, we are optimizing around hybrid flash implementations. We have a new set of Tiering 2.0
on Gen6. All Flash Arrays (AFAs) must have dedupe and compression to be competitive. In MSA’s hybrid
tiering world, there is not much performance benefit to compressing and deduping the SSD tier. The
work required to compact the data and then uncompact it just don’t balance out. It is more work than
the possible benefits. Bottom line - MSA is not an AFA. Sell AF20 Nimble for that.
Question There has been no reference to all-flash configs. Is this not recommended?
Answer: As mentioned above, successful AFA's need Compression and Dedup. MSA continues to evolve and
optimize the Performance Tiering engine for Hybrid efficiency. MSA users can configure all flash
configurations and they do. There are Reference Architectures out there for SAP HANA configs which
only use SSDs. This is a unique configuration where there is an ability to take advantage of MSA’s raw
horsepower in these configs….it is not the standard. If a customer is looking for a general purpose AFA,
Sell Nimble or Primera.
Question: Are the MSA SED solutions considered Data at Rest encryption?
Answer: SEDs drives for MSA do provide Data at Rest Encryption. All drives must be SED to do encryption. We
continue to sell SED drives and non-SED or regular HDD/SSDs.
Answer: No, MSA Encryption is accomplished SED HW based encryption capabilities in the HDD/SSDs. SW-based
encryption is not offered or supported.
Question: Are there any plans to offer a Centralized Key Management system?
Answer: No. There are no plans to qualify or develop any External or Centralized Key Management system for the
MSA Full Disk Encryption solutions.
Question: Can we please do better documentation of the cabling with this release?
Answer: Backend cabling has been simplified to only using the straight thru cabling with Gen6. Host side cabling
is recommended to utilize the same ports on both controllers but with the move to explicit only
mapping and mapping to all ports the host side cables can be connected to any port on the opposite
controller for redundancy. Please let us know where you are seeing difficulties so we can help make it
better.
Question Is it mandatory to populate SFF slots in an MSA 1060, or can we just add the 3 LFF shelves and only
populate them?
Answer: That could be done but the configuration does not make much sense financially, the delta between MSA
1060 and 2060 is not much difference when compared to a LFF 2060. Take a look at the math.
Question: Has performance measurement and reporting, especially history reporting, improved over the existing
units?
Answer: Yes. The new SMU v4 GUI is far better than previous version. Performance monitoring and dashboard is
far advanced to v3.
Question: VRO/QLC SSD drives haven't been mentioned. What is MSA’s plan as far as support for these new
SSDs?
Answer: We are looking in to it for sure. AFA arrays are first to pick it up. We would see it as a potential
replacement for 10k HDDs, however the cost math is not really lining up. It will at some point and we
will productize Gen6 SSD media.
Question: MSA Gen6 has new SMU v4. What is special about it?
Answer: The SMU v4 is has been rewritten from the ground up. The Dell ME4 is still on the old one. The new GUI
worked very hard to improve and modernize the dashboard so you can get a great “at-a -glance” view
with abilities to drill in. It is MUCH better thanv3 SMU. We now have accurate performance reporting
V3 – September 2020
and also a historical change log. It looks MUCH more modern and fresh visually. Second big area of
improvement is the use of “guided workflows” to help a user do complex tasks that they might only do
once in a long while. Examples – Creating Disk Groups and Volumes, initial unit setup. By guiding the
user through the workflow, we can eliminate making mistakes, errors, skipped steps and needing a
manual. GUI guides the user through the tough stuff.
Question: What is the min number of drives required to start the array?
Answer: 2 drives (RAID 1) is the minimum configuration.
Question: Do we have to use MSA-DP+ or are the old RAID types still available?
Answer: MSA-DP+ is highly recommended for HDDs, but other levels of RAID are supported as well.
Question: Please clarify the "we don't recommend this (MSA-DP+) for SSDs" comment. I assume this is a
performance related issue, but real customers are often willing to trade top-end performance for
manageability, especially in the SMB space.
Answer: It is absolutely NOT a performance issue. It has to do with the fact that MSA-DP+ requires that you start
with 12 drives. Very few MSA customers require so many SSD drives upfront. Additionally, SSDs do not
commonly fail and the resiliency MSA-DP+ provides would not be as advantageous as with HDDs.
Question: Does this mean that the "power of 2 method" of placing disks in a disk group falls away?
Answer: Yes and no. With MSA-DP+ the stripes will be configured to follow the Power of 2 conventions without
having a power of 2 # of drives in the disk-group. The minimum of 12 drives makes it so under the
covers. For any other RAID types, Power of 2 rule still applies.
Question: For MSA-DP+ disk groups, the type of disk must be the same (though different in capacity) or can it be
different disk type?
Answer: Correct, the MSA-DP+ disk group will conform to the tiers for the drive types (SSD, SAS, MDL-SAS) but
you can have an MSA-DP+ disk group on each of those tiers. 1x MSA-DP+ disk group of SAS drives and a
different disk group of MDL-SAS drives in a different disk group.
V3 – September 2020
Question: With the past generations of MSA the maximum number of disks within a group was 16 or 32 disk (if
you configure RAID 50) With the MSA-DP+ it is possible to make groups bigger than 16/32 disk
independently of the raid?
Answer: MSA-DP+ should be treated as a RAID level. You can have between 12 and 128 drives in an MSA-DP+
disk group. RAID 50 has not been supported on virtual storage since its introduction on Gen 4 arrays.
Only the now obsolete linear disk groups supported this and is no longer required to achieve the same
benefits (wide-striping).
Question: MSA-DP+ stripe zones have a minimum drive quantity of 10, but you can grow it one by one?
Answer: The minimum number of drives is 12 for an MSA-DP+ disk group due to the included spare capacity. The
disk group can then be expanded by 1 or many drives. The disk group will rebalance the stripe zones
when a new drive is added.
Question: In this case, customers have to start with 12 HDDs per pool?
Answer: To use MSA-DP+ disk group the customer will need to have a minimum of 12 of the same type of drive
(SSD, SAS or MDL-SAS) to create the MSA-DP+ disk group. We recommend MSA-DP+ for HDD tiers, but
other RAID types that require fewer drives (e.g. RAID 6) is also supported. However, we do not
recommend RAID 5 for HDDs.
Question: Will this performance impact/rebuild time table be available in a white paper somewhere once the
Gen6 announcement is made?
Answer: This is documented in the sixth-generation virtual storage technical reference guide. MSA-DP+ is big
going forward.
Question: Can you possibly share some basic calculation method to calculate usable capacity until a sizing tool is
available?
Answer: This is especially for MSA-DP+ it is absolutely coming - Sizer, examples, best practices.
Question: If I want to use both controllers (2 pools) and MSA-DP+ with tiering, what’s the minimum number of
disks?
Answer: 4x SSDs and 24x SAS. Each Pool would have 2x SSDs (R1 pair) and 12x HDDs (MSA-DP+ group)
Question: Is the flash to disk ratio based on RAW or Usable capacities of the capacity tier?
Answer: Useable
V3 – September 2020
Question: So, are you seeing (or advocating) MSA configs where all drives and SSDs are run by just one
controller, with the 2nd controller as a standby controller, similar to how Nimble runs? Your
comments about SMB customers not exceeding the performance of one controller seem to point to
this scenario.
Answer: Yes. With the controller performance available today MSA no longer requires Dual Pool (active active)
for most SMB customers. You can get a LOT more performance than most SMBs need with one Pool
(Active Passive) configs like Nimble does. However, this options remains available when very high
performance is required and lowers the $/IOP ratio.
Question: In the future, will the HPE MSA integrate with StoreOnce and RMC?
Answer: The POC is already done with RMC/MSA Gen6. We are scoping the release mechanisms. It would give us
data mobility in HPE ecosystem and application awareness for our replication services. It will take some
time, but we would like to be on the next big RMC platform release.
Question: Will SAF tools be able to collect performance data from this generation product?
Answer: No SAF integration, use and understand the I/O Workload tool within the capacity section of the
new GUI.
Question: Is there a way to size and MSA before the new tool is released?
Answer: Yes. We plan to make a simple Capacity sizing tool to the field which will facilitate figuring out useable
capacity and configurations on the new Gen6 platform. To accompany this simple tool, we will release a
PPT with a number of simple but common examples. For those we will include the performance data
curves which come from the PEM (Performance Estimation Module) datasets the Ninja-based tool will
be pulling from. As a final tie-in, we will do a TekTalk-on-Point which will walk through these examples.
When the official Ninja.next tool is released, we will follow up with an official TToP on how to
size/configure using that too.
Question: Will we detail the R/W ratio and I/O sizes along with the 350K IOP numbers?
Answer: The performance numbers for our “hero” specs will be included in the QuickSpecs as they have been
with previous generations of the MSAs. Details of the tests are included along maximum Random Read
and Write numbers as well as Sequential.
Question: Are those IOPS only considering backend? Or are there new metrics that take controller performance
into consideration?
Answer: The 325k IOPs are as measured from the host in an end-to-end configuration.
Competitive
Question: Who else is OEM’ing these same boxes from Seagate? Will Lenovo have the same product?
Answer: Dell purchases a very similar solution from Seagate (ME4). It is a half-generation behind MSA Gen6. We
have the newest acceleration ASIC, new GUI, RESTful interface, Tiering 2.0, MSA Health Check, and a few
other things. Lenovo used to purchase a G4/Gen5 product from Seagate but they have dissolved that
relationship and moved to a new joint venture with the LSI side of NetApp.
Question: Lenovo has a NetApp OEM offering that is very aggressive. Nimble HF20H is good but has limitations
from storage capacity upgrades. Will MSA compete against this?
Answer: Yes, MSA will compete with the lowest end offerings. Lenovo seems to continue to thrash as to Entry
Storage Strategy year to year. They were taking IBM’s v3700, then the OEM'd from DotHill for a couple
years, now they have this new Joint Venture with NepApp...these shifts to leave customers hung out to
V3 – September 2020
dry. The JV with NetApp is new so no clue if it will work out in long term. Should be easy enough to sell
against. It is not just a price game remember.
Positioning
Question: What is MSA’s position in the HPE Primary Storage Family?
Answer: Position MSA where the budgets are $20k or less. From a Host connectivity standpoint, MSA ships with
FC, iSCSI or SAS connectivity. FC is ~50%, iSCSI 25% and SAS is 25%. MSA 1060 series is designed for
lowest starting at price point. MSA 2060 is the building block flexible config array and the MSA 2062 is
the Hybrid Flash model from day one.
Question: Are there specific workloads or deployment methods where MSA is perfect? How about not so
perfect?
Answer: MSA is best positioned as a “general purpose” shared storage arrays. It is run in virtual deployments a
high percentage of the time. For SMBs using an array like MSA, the array is expected to run ALL VMs
(applications) it takes to run the business. MSA is also sometimes selected to run on application (like SAP
HANA) and optimized for speed. The other common deployment mode for MSA is that of Branch sites
(ROBO)…..the more of them the better for MSA. This deployment method really depends on optimized
costs per site ($$/Site).
Question: Does HPE have a low cost, general SAN Software Defines Storage (SDS) solution now that StoreVirtual
is end of life?
Answer: Yes. StorMagic is sold via HPE Complete.
Question: Should MSA and File Controllers still be positioned for mix File/Block requirements?
V3 – September 2020
Answer: Standard (low performance models) are already EOL announced but they are still shipping. Performance
models are still shipping. They will be going away as the Primera/Nimble customers will have a new
strategy going forward. While HPE has been using the File Controller strategy, we have seen sales of the
low-end General File Controller has dropped to very near zero. For MSA, we are working on some
solution whitepapers on how customers deal with this through hosted VMs and/or StoreEasy
appliances. Many customers are already doing things this way today.
Question: I see that you have StoreEasy in the slide. With Microsoft apparently discontinuing the Windows
Storage Server on 2019, what is our strategy going forward?
Answer: Our StoreEasy platform will continue as it has regardless of Microsoft's licensing strategy. Customers
who want a turn-key appliance will always have StoreEasy option. If they want to roll their own file
server, they don't want/need an appliance. Simplicity is at the heart of StoreEasy appliances.
Question: So I've heard, however I run into a lot of customers who do not want to deal supporting both
hypervisor resources and OS updates for their File environment.
Answer: For mid-range, high end this is a common ask. Coming up far less frequently in <$20k space. Customers
often host a file system VM on the block array. Not worth having the HW dedicated just to file. Make it
fast for block and optimize there. Don't sacrifice on block for just little amount of file requirements.