HP P4900 SSD Lefthand – Bloggers podcast

A few weeks ago HP launched their new kid on the ‘block’, the P4900. It is an SSD-only box in the P4000 (Lefthand) family. There are two models: 3.2TB (single node) and 6.4TB (dual node). A single node contains 8x400GB MLC SAS SSD’s. As we know SSD solutions still have a lots of open questions in the community. At the launch of a new product the vendor will always show you how good the product is. This is good because we know at that point how much more it can do for our environment.

But knowing how much more you can do also has constraints. And those are the questions/answers you will not find that often in those marketing campaigns. Therefore Calvin Zito  invited some bloggers including myself for a podcast together with the product specialists.


On the podcast:
Host: 
   *   Calvin Zito@HPstorageGuyblog
HP: 
   *   Kate Davis@KateAtHP – 
Worldwide Product Marketing Manager P4000

   *   Dale Degen – Worldwide Product Manager P4000
   *   Charles Randall – P4000 Engineering team
Bloggers: 
   *   Chris Evans@Chrismevansblog
   *   Luigi Danakos@Nerdblurtblog
   *   Hans De Leenheer@HansDeLeenheerblog


Capabilities versus constraints?

  • The advantage of the P4000 concept (1 node = 1 controller) enables us to use more SSDs in a cluster that would otherwise be possible in a classic 2 controller system (i.e. EVA) because of the CPU constraints of those systems.
  • Due to that clustering with a fixed amount of disks per controller, the group performance scales linearly.
  • Although you would theoretically be able to scale an unlimited amount of nodes, HP gives us some best practice numbers. From a Cluster perspective HP advises to go up to 10 – 16 nodes and from a Management Group perspective max 32 nodes.
  • You can have multiple types of disks in a Management Group but not within the same Cluster. This means on one hand that you can change LUNs “on the fly” from 10k to 15k or SSD but on the other hand that this implies the entire volume (no sub-lun tiering)
  • One of the greatest advantages of the P4000 technology is the active/active multi-site SAN. Be aware that with the P4900 the distance between the 2 sites will need to be shorter than with spinning disks. This is due to the fact that every IO has to be written to both sites before it can be acknowledged (round trip time latency).
  • 10GbE is necessary! This is not for the server>san connections but for that backend IO between the nodes that is necessary to get the IO acknowledgement on time back to the server.
  • If you look at the 2 models available now (3.2 <> 6.4) be aware that there is no difference between the two but that the 3.2 is only a single node, single controller (hence SPOF!). – UPDATE: HP tells me that the 6.4 is the base model and the 3.2 is the upgrade node to that. 
Future Use:
  • I’d love to see HP implementing Sub-Lun tiering in the midrange products (P4000/P6000). 
  • I’m also looking forward to the 15k SAS on Gen8 with the “Dynamic Workload Acceleration” flash enhancements and a P4900 SSD box on SCSI-Express motherboard. What I’d like to test now as these products do not exist yet is a Gen8 server with the flash enhancements, 15k disks and a P4000 VSA versus the current P4500.
MyThoughts:
Good stuff as is the whole system around the Lefthand family. If you know the constraints I would definitely find enough use-cases where this fits in. Think VDI environments or Scalable Databases but also burst capacity (with peer motion) for temporary batch workloads or development systems.

Others Links:
Exclusive announcement: SSD-based P4000, the P4900
HP P4000 Lefthand Peer Motion DEMO
P4000 Zero to VSA DEMO

Trackback Links:
HP – Around the Storage Block – the official blog about this podcast
HP – Storage Blog – official blog pointing to official blog about this podcast – #lol

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.