Before I get into why I think Pensando is changing the game one card at a time, let’s first talk about who they are and what they bring to the table.
Pensando Systems is a pioneer in distributed computing designed for the New Edge, powering software-defined cloud, compute, networking, storage, and security services to transform existing architectures into the secure, ultra-fast environments demanded by next-generation applications.
So what makes these cards so cool?
Well, they have a set of programmable P4 processors that allow it to hold multiple personalities. These personalities can be used to provide offloading of numerous services that can task an x86 processor. But why is this important?
Let’s take a quick walk down memory lane to find out.
Back in the day, many companies used offloading modules to alleviate some of the CPU heavy lifting. The CPUs at the time just didn’t have the horsepower at the time due to Moore’s law. The 80×86 was one of the 1st CPU’s to offer a math co-processor option, Cisco’s claim to fame was the use of ASIC’s to dominate the router market. Both the math co-processor and ASIC’s allowed the Operating Systems at the time to take full advantage of the hardware to be as fast as possible. Fast forward to super high clock speeds and multi-core CPU’s era and we get companies like Arista that became blazing fast by not utilizing specialized modules like an ASIC.
What did we gain by catching up to Moore’s Law and having super-fast CPU’s?
Well, it paved the path for Virtualization via products like VMware and KVM to consolidate many servers onto fewer larger ones. Next came Hyper-Converged Infrastructure systems from NetApp/HPe/Nutanix to name a few and Software-Defined Services like Storage (vSAN) and/or Networking (NSX). Which also gave way to the hyper cloud scalars we know today (AWS/Azure/GCP/etc).
So much innovation in this space, but can we do better?
Enter Pensando with their Capri cards. These cards can offload a ton of processes from x86 CPU’s, some examples include Network, Security, and Storage. Let’s look at them one by one and which companies/systems could possibly utilize these cards.
Network Services
I believe almost all companies can leverage offloading network services of one type or another. As basic as adding multiple speed options from 10G – 100G all the way to protocol routing/etc. Here companies like Arista/Cisco/Extreme/etc can utilize these cards to help speed things up with very low latency, high bandwidth offerings from these cards. HyperScalers I think will benefit the most from the use of these cards. Not only are you speeding things up, allowing for the servers CPU be used to provide VM hosting cycles but by consolidating services you also lower the TCO of each server.
An example could be instead of just hosting let’s say 20 VM’s that act as NAT Gateways, you might be able to host now 100 NAT Gateways and use the remaining CPU cycles to host other VM’s that you can’t easily offload to the card yet.
Security Services
Most firewall and/or security-focused systems made by Cisco/Fortinet/Palo Alto/etc can also benefit from utilizing these cards.
With the need to inspect a greater amount of traffic these days why not consolidate again here with not only utilizing the card to provide network connectivity up to 100G but also offload DDOS attacks which ultimately is targeted to cripple the CPU of the device in question. This will allow the CPU to run the OS and other monitoring subsystems needed to fight the attack.
Another possible example could be to install these cards in an HCI solution. Wait that’s crazy talk?!? Think about it, if we could get NSX to offload these services from management VM’s and onto cards that we need no matter what to provide network connectivity, we free up much-needed CPU resources to host real application VM’s.
Storage Services
Here is where things get pretty cool. Imagine offloading from the CPU all the normal services like:
-
- Compression/Decompression
- Compaction
- Deduplication
- RAID coding
Enable Network transportation as well as encryption and you have a killer combination. Why do I mention this? Well, NetApp and Pensando have already made this happen in the A400 line of systems. Other storage companies like HPe/Dell EMC/Pure/etc can also leverage these cards in their systems to bump performance and lower TCO.
Final Thoughts
I believe there is a great opportunity for companies that provide HCI systems to greatly benefit from the consolidation of subsystems and network IO functions on to Pensando cards while also lowering the need for a controller or special purpose VM’s to run systems like NSX & vSAN giving customers much-needed CPU cycles back by eliminating the VM tax.
The next group which will benefit from all 3 profiles offered by Pensando are the Hyperscalers. By consolidating all offerings they can provide a greater amount of CPU cycles to their customers and lower their TCO which is a win-win.
Reviewing the Storage Field Day presentations over on: https://techfieldday.com/event/sfd20/ and specifically these 2 presentations have inspired me to write this review. I encourage everyone to view the following videos for more information.
You must be logged in to post a comment.