SAP HANA on Intel Optane PMEM
Key Points
- Bradley Knapp, an IBM Product Manager, explains how Intel Optane DC Persistent Memory (PMEM) can be used to host SAP HANA databases.
- PMEM is a NAND‑based DIMM that sits between DRAM and NVMe storage, offering much higher speed than SSDs at a lower cost than RAM, thus filling a performance gap in the storage hierarchy.
- IBM Cloud servers with PMEM come pre‑configured in App Direct mode with namespaces applied, so customers can immediately begin a HANA installation.
- In a HANA deployment, the columnar store (about 95% of data) is placed on PMEM for persistence, while the volatile row store (roughly 5% of data) remains in traditional RAM for ultra‑low latency.
- This architecture enables greater system consolidation and performance benefits without requiring an all‑RAM configuration.
Sections
- Optane PMEM for SAP HANA - IBM Product Manager Bradley Knapp outlines what Intel Optane DC persistent memory is, its position between DRAM and SSDs, and how it boosts performance for SAP HANA databases.
- PMEM Enables HANA Consolidation - The speaker explains that using persistent memory in App Direct mode lets multiple HANA column‑store databases share one physical system, dramatically reducing footprint and total cost of ownership while preserving data across power cycles to avoid costly reloads.
Full Transcript
# SAP HANA on Intel Optane PMEM **Source:** [https://www.youtube.com/watch?v=S0KanFX2KT0](https://www.youtube.com/watch?v=S0KanFX2KT0) **Duration:** 00:04:40 ## Summary - Bradley Knapp, an IBM Product Manager, explains how Intel Optane DC Persistent Memory (PMEM) can be used to host SAP HANA databases. - PMEM is a NAND‑based DIMM that sits between DRAM and NVMe storage, offering much higher speed than SSDs at a lower cost than RAM, thus filling a performance gap in the storage hierarchy. - IBM Cloud servers with PMEM come pre‑configured in App Direct mode with namespaces applied, so customers can immediately begin a HANA installation. - In a HANA deployment, the columnar store (about 95% of data) is placed on PMEM for persistence, while the volatile row store (roughly 5% of data) remains in traditional RAM for ultra‑low latency. - This architecture enables greater system consolidation and performance benefits without requiring an all‑RAM configuration. ## Sections - [00:00:00](https://www.youtube.com/watch?v=S0KanFX2KT0&t=0s) **Optane PMEM for SAP HANA** - IBM Product Manager Bradley Knapp outlines what Intel Optane DC persistent memory is, its position between DRAM and SSDs, and how it boosts performance for SAP HANA databases. - [00:03:07](https://www.youtube.com/watch?v=S0KanFX2KT0&t=187s) **PMEM Enables HANA Consolidation** - The speaker explains that using persistent memory in App Direct mode lets multiple HANA column‑store databases share one physical system, dramatically reducing footprint and total cost of ownership while preserving data across power cycles to avoid costly reloads. ## Full Transcript
Alright, thanks for stopping by today.
My name is Bradley Knapp, I'm a Product Manager here at IBM, and I want to talk to you today
about how to use Intel Optane DC persistent memory, or PMEM, for SAP HANA. So, the actual HANA database itself.
Intel and SAP cooperated very closely in the development of PMEM and the development of
HANA to allow you to put it on top of PMEM, and the ability to do so was actually introduced
in HANA 2.3 back a couple of years ago. And so, before we get too far into that I did want
to take just a second let's talk about what PMEM is, what this Optane DC persistent memory
product is. So, Optane DC persistent memory is a NAND chip, right? So, it's a persistent chip
similar to what you would find on an Optane SSD or an Optane PCIe drive, but it's in a
DIMM form factor. So it's on the the actual piece of silicon that goes into a DIMM slot
just like RAM does. Why is this important?
Well, if you think about our storage pyramid, right, where we've got tape at the bottom, and then we've
got hard disks, and then we've got normal SSDs and PCIe drives - the higher the throughput
and the lower the latency between the processor and the storage, you're gonna see increased
performance, right? Performance always goes up, and so PMEM occupies an interesting spot
between DRAM, which is the lowest latency but the highest cost, and NVMe storage and PCIe drives.
And so, by filling that gap where PMEM is orders of magnitude faster than a normal SSD, but less
expensive than RAM.
It really allows you to do additional system consolidation that wouldn't be possible if
you were running on a system with just RAM. And so, whenever you order a server that's got
PMEM on it from IBM Cloud you're gonna receive something that's set up, it's ready to go, it's
fully provisioned, and it has everything you need to go ahead and start installing SAP HANA on it.
And so, that means that the, that the PMEM is already set in app direct mode, that it's already
had the namespaces applied, and so really you can go in and start that HANA installation. Alright, and
so now that we're ready to actually install that HANA database we need to think about a
little bit about, “how do we do that installation?” - and the answer is HANA takes care of all of
it for you. Remember, the whole idea behind HANA is that you've got a columnar database
that's running in memory-and remember those columns, that column data store, that's where
95% of your data is - and that column store is what we're going to put into the PMEM.
That's what's going to run in the persistent layer. The volatile piece, the volatile data structures, which
is the row store, that's going to continue to run in RAM, just like it always has, because
you need that lower latency and you need that a higher level of access, but remember that
volatile piece, the row store, that's only 5% of your data.
And so, you're gonna put that 5% on the RAM, the 95%, the column store, that goes onto
the PMEM, and then just like always your log array, your data array, and your shared array, those go
on traditional disk, right?
You're still gonna write those down, but the benefit of having all of this extra space
available in the PMEM for your column store, means that you can use technologies like MDC or MCOS to
put additional HANA databases on to this same physical system, moving all of those additional
databases onto the same system makes it possible to decrease your total footprint size,
which is going to bring down your cost. You're actually consolidating multiple physical machines onto
a single machine, and the benefits of that is it's gonna dramatically decrease your
total cost of ownership.
And then, in addition to the to the decreased cost that you get by doing that consolidation, let's
also not forget that the PMEM itself is persistent, right? The data, the column store that's residing on it, it
survives through a power cycle - so if you have to do a system restart as part of your update
cycle, or anything like that, you don't have to go through that very painful process of
having to reload all of your tables from the data store back up into memory again.
They're already there because they were running on that persistent layer and got you through
the power cycle. So, that's kind of an overview of how you would put HANA on top of PMEM in
App Direct mode.
If you have any additional questions, like always please drop us a line below.
If you want to see more videos like this in the future,
please do like and subscribe, let us know about it.
And don't forget you can always get started on the cloud at no cost
by sending up for a free IBM Cloud account by going to cloud.ibm.com.