The Dell EMC High-End Systems Division talked about two systems. First about the VMAX All Flash, and later about the XtremIO X2. This post is about the latter one. The XtremIO X2 builds upon the foundation of the original “old” XtremIO, but also does a couple of things differently. This post will explore those difference a bit, and will also talk about asynchronous and synchronous replication.
Dell EMC XtremIO X2
The XtremIO X2 improves upon the original XtremIO in several key areas.
For starters, there’s more flexible hardware scaling. XtremIO X2 comes in two flavor of X-Bricks: X2-S and X2-R. The X2-S can start smaller at 7.2TB raw capacity and scale up to 28.8TB raw, while the X2-R starts at 34.5TB and can scale up to 138.2TB raw. These are capacities per X-Brick, and an XtremIO X2 system can can scale up to 4 (X2-S) or 8 (X2-R) bricks. There’s no mixing of X-Bricks though, so check your data growth before you purchase a system.
While we’re talking hardware: you can now add drives to an X-Brick. Each X-Brick starts with 18 initial drives, and can then scale with 6 drive increments up to 72 drives. The DAEs are top loaded, as you can see in the picture on the right.
Another fancy hardware improvement: there’s fewer cables. The first time I spotted an XtremIO X1 system in the wild, I was shocked by the amount of cables. In the X2, the BBU is gone (replaced by a PCIe NVRAM unit with a super-cap). As a result, cabling has been reduced.
Software improvements have enabled the XtremIO X2 to achieve a better data reduction of 5:1, compared to the X1’s 4:1. This means more bang (=TBs) for your buck. Partially this might be because the X2 uses 16KB blocks compared to the 4K/8K for the X1 XtremIO.
It’s interesting to see the move to 16K for XtremIO, from 4K then 8K. Tradeoff of granularity & savings from things like compression #SFD14
— Chris M Evans (@chrismevans) November 8, 2017
And don’t forget the HTML5 interface, which became somewhat of a common topic for all presentations at Storage Field Day 14. This not only saves the engineer from installing 12 different Java versions and browsers. It also means that customers just need a regular browser in their virtual desktop image, without potentially insecure add-ons!
Replication
The XtremIO X2 comes with native asynchronous replication that does not require any WAN accelerators. Hashes are sent to the destination system in batches and matched. The actual data is only sent over if the hashes are new to the destination system. This saves immensely on WAN link bandwidth. The added latency doesn’t really matter since it’s asynchronous replication anyway.
This sparked a discussion that continued later in the car: is synchronous replication still as relevant as it used to be? There are two apparent trends:
- Media and thus storage systems are becoming increasingly faster at lower latencies. Pretty much all systems now achieve <1ms latencies, and many more achieve <0.1ms. In fact, storage media is no longer the bottleneck. Companies like Intel are doing fancy things to improve software and controller latencies to reduce end-to-end storage latency even more. If you’re adding synchronous replication, WAN latencies will start taking up relatively more of the overall latency.
Good observation from @toddtoles: are we still in a synchronous replication era? With array latencies of <0.1ms, do you really want to add WAN latency that's multitudes of milliseconds? #SFD14 @DellEMC
— Jon Klaus (@JonKlaus) November 8, 2017
- We used to default to storage array replication for everything. However there’s a (partial) move back to application based replication. Or the replication is provided by a different, bolt-on solution because a customer also wants VCR-like continuous data protection. Something that RecoverPoint offers for example…
My thoughts on XtremIO X2
First of all: impressive and predictable performance. Due to the “inline everything” design, there are no performance drops due to garbage collects or post-processing deduplication. Our company has got a couple of the X1 systems in production and so far we’re getting the promised data reduction and performance.
It’s also easier for new customers to start with XtremIO: X2 X-Bricks can start out partially populated with disks. This enables customers to start small and scale it up and out when the need arises. Read: the applicable market for XtremIO X2 just got bigger.
Todd did have a very valid point that even a very fast WAN latency of 2-3ms would wreak havoc on the XtremIO overall latency with synchronous replication. And I’m seeing many customers moving away from synchronous replication in favor of “near RPO=0” replication. However I can envision multiple scenarios where replication is on-campus and thus latency isn’t as much of a problem. In the end it’s customer demand that will drive the replication type, and thus whether XtremIO will feature synchronous replication in the future.
Check out the rest of the presentation (and especially the copy data management which I didn’t cover) over here.
Disclaimer: I wouldn’t have been able to attend Storage Field Day 14 without GestaltIT picking up the tab for the flights, hotel and various other expenses like food. I was however not compensated for my time and there is no requirement to blog or tweet about any of the presentations. Everything I post is of my own accord and because I like what I see and hear.