My Takeaways from the Pure Storage FlashArray//m Announcement

I’ve been paying close attention to Pure Storage’s announcement yesterday. There was significant marketing hype around this news, and I was curious what one of our biggest competitors had to show. Pure Storage announced a new FlashArray//m all-flash array family with capacity and performance improvements, updated cloud and analytics-based support, as well as an enhanced upgradable business model.

Pure1, the cloud and analytics support facility, was already presented back in September 2014 at Storage Field Day 6. Evergreen is a natural progression of the Forever Flash upgradability business scheme. These two items are a good natural progression of Pure Storage’s strategy, but they are nothing new, and they are not game changers in the value the company brings to the market.

The more significant aspect of the announcement is FlashArray//m, which is basically a hardware refresh. There are three FlashArray//m models – the m20, m50 and m70. The new arrays are replacing the old FA-4XX models, which according to Pure can be upgraded to “m” status non-disruptively. The refresh contains a more up-to-date processor and networking technologies, a very natural progression. The more notable advancement is the use of NV-RAM modules replacing the SLC SSDs that were used as “NVRAM” in the FA-4XX models and were significant hardware bottlenecks for write IOs. Write cache with memory speed is actually a standard in the industry. Kaminario has been using it for the last few years and Pure Storage made a good move to follow this direction (for Pure, it is actually near memory speed as the interface to the device is PCIe).

The main motivation for the FlashArray//m hardware refresh is to meet real customer requirements to better scale capacity and performance in a cost-efficient way. The old FA-4XX family was not scaling well in both capacity and performance compared to other all-flash arrays. As such, there were two possible strategies to solve this issue:

  • A significant change to Pure’s existing active-passive scale-up architecture, to also support scale-out with a real active-active design.
  • Incremental performance and capacity improvements by refreshing hardware with the existing scale-up architecture.

Pure Storage decided to take the easier path and refresh the hardware rather than tackle the core architecture challenge. They gained a nice 50% performance improvement and density improvement utilizing SSDs that are available in the market (Kaminario is planning to use new higher density SSDs this year).

When Pure’s CTO John “Coz” Colgrove was asked why they do not support scale-out, he said that scale-up is more cost-efficient and that 98% of the market should be satisfied by the performance numbers of the m70 model. He also said that scale-up is not a religion and they might add scale-out in the future.

Let’s tackle both claims. I agree that the maximum performance specifications of 300k IOPS and 9GB/seconds is impressive, but is this truly the load that’s relevant for the majority of customers? The statistics we gather in Kaminario’s Health Shield indicate that 35-40% of the IO is write IO and for a significant portion of our customers write IOs are above 60% of the entire load. In practice, we see that high write load is the primary performance motivation for scaling out the cluster to multiple nodes, and in many cases, is a deal breaker when we POC against Pure Storage. As a customer, I would ask Pure Storage to be more open and share the write IOPS and bandwidth with 50% and 100% write IOs.

Regarding the claim that scale-up is more cost-efficient then scale-out, I totally agree with Coz. There is room in the market for both scale-up and scale-out storage systems. In some scenarios, scale-out requires adding unnecessary costs for the customer, and the extra cost limits the adoption of scale-out storage systems. In other scenarios, the performance and capacity requirements can be fulfilled only with scale-out architecture. The conclusion is that both scale-up and scale-out capabilities are critical for any modern and flexible storage system.

Kaminario’s K2 is the first storage solution to close this gap by enabling both scale-up and scale-out within the same system, allowing customers to scale and meet performance and capacity requirements in the most cost-efficient and flexible way. You can read this article for more information.

Oh, and one more thing… it is not a surprise that I really like the new shiny bezel as it is an orange version of the K2 bezel that has been illuminating our data centers for the past two years :).

phpAOobeQPM

New Call-to-action