Synology launched a trio of enterprise hard drives at 8TB, 12TB, and 16TB capacities along with a trio of “enterprise” NAS units. As part of the dual announcement, Synology is locking its new NAS units from using large hard drives from other vendors, instead only allowing its own hard drives to work in the units. Let us get into what is happening.
Although we do not normally do videos for news, we have been getting a lot of questions about this, so we have a video:
As always, we suggest watching that in a YouTube tab. That video is meant to be listed in the background.
Synology HAT5300 Series
Synology is not manufacturing hard drives. Instead, the new Synology HAT5300 series is effectively made from re-branded Toshiba drives. There is an 8TB model at $239, a 12TB model at $349, and an upcoming 16TB model that does not have a price or release date yet. As part of the offering, Synology says it has done some firmware work on the drives to help increase performance. Indeed, Synology says the HAT5300 drives deliver up to “274MB/s sustained data transfer” and are up to 23% faster in sustained sequential read performance in demanding multi-user environments compared with other similar-class drives. Here is the performance testing footnote:
Performance testing was conducted by Synology using 12 drives on an SA3600, configured using RAID 5, against similar class drives (Enterprise) with IOMeter (64KB blocks). Results are for reference only. Actual performance may vary depending on workload, testing methods, and how devices and software are configured. (Source: Synology)
That performance note makes the 23% claim effectively useless since Synology is not naming the capacity nor drive they are using for comparison. A Western Digital HC550 18TB drive is both bigger and has a sequential transfer speed of 269MB/s and 252MiB/s. The Seagate EOS X18 drives can also hit 18TB and are rated at 270MB/s sequential transfer speeds. Again, since Synology’s performance note says IOMeter 64KB is being used along with MB/s, not MiB/s, the 23% faster rate assumes competitive drives operate only at 222MB/s. It seems Synology is using older generation drives as a comparison point.
The move by Synology to release its own hard drives makes sense from a corporate perspective as it is a good way to show its investors continued revenue growth. The cost of the bare NAS itself is often half or less of the total system.
Synology Trio of New Enterprise NASes
Synology also has a trio of new NAS units that it is calling an Enterprise line. These NAS units do not necessarily have features such as Active-Active controllers (Synology has Active-Passive using external connections and software), dual-ported SAS drives, and other similar features. Instead, we think these units seem more of an attempt by the company to reach beyond the SMB space and into the SME space for the low-end of the enterprise market.
The first unit is the Synology RS3621RPxs. This is a 2U 12-bay unit:
This unit has an Intel Xeon D-1531 (6C/12T) processor, 4x 1GbE networking, and an 8GB ECC UDIMM for memory.
The second unit is the Synology RS3621xs+ which is a similar 2U 12-bay unit:
Here we get an upgrade to an Intel Xeon D-1541 8-core processor. This unit also adds dual 10Gbase-T networking as standard and keeps the 8GB ECC UDIMM. Memory capacity, like the other units in the trio, is listed as 64GB maximum.
Finally, we have the Synology RS4021xs+ which is the 3U 16-bay option:
Here we get the same Xeon D-1541 and dual 10Gbase-T that we saw on the RS3621xs+. Aside from adding four more drive bays, this unit also upgrades to a 16GB ECC UDIMM.
Discussing the Specs: Infiniband
First off, we want to field a question we have received on these units multiple times: Infiniband. Listed on the spec pages for all of these units is 2x Expansion Ports (Infiniband).
The Intel Xeon D series was launched in 2015. Intel Xeon D – Intel SoC Changing the low end with Broadwell-DE and Synology is using the post-Xeon D-1500 series and SR-IOV fixed variants of the chips with its new NAS units. Given the other I/O that the company is showing including dual PCIe 3.0 x8 slots, it is unlikely FDR/ EDR/ HDR Infiniband that could be repurposed into a high-end RDMA fabric.
Instead, we think that what Synology is actually stating is that it uses a CX4/ SFF-8470 connector. This was a popular connector for Infiniband cables 10+ years ago. These are labeled as “11 Expansion Ports” on the units above. These ports are meant to connect to 1-2 disk shelves.
In the enterprise storage space, Infiniband has been used to connect multiple controllers together and also storage to servers. Indeed, Infiniband is often associated with high-performance storage today, but it started as a storage interconnect many years ago.
We hope that Synology clarifies this in its documentation.
Discussing the Specs: Xeon D and Networking
The Intel Xeon D-1500 series originally came out in 2015. After the SR-IOV issue effectively forced an updated chip from the launch Xeon D-1540 and D-1520, the company released new models, higher core count models (up to 16 cores publicly), and models with more networking. After the first generation Intel Xeon Scalable launch, we saw the Intel Xeon D-2100 series launch to bring the new architecture to the embedded market. We covered this evolution in exploring Intel Xeon D Evolution from Xeon D-1500 to Xeon D-2100. We also covered how the Twin Lakes Intel Xeon D-2100 platform powers Facebook. When the 2nd generation Intel Xeon Scalable series launched in 2019, Intel updated the D-1500 series with new higher clock speed parts in its Xeon D-1600 update.
Synology is using the Xeon D-1531 and D-1541 in its three new NAS units. These are Q4 2015-era parts. For new NAS units launched in 2021, that means Synology is still using current-generation parts, but they are certainly older.
For some reason, Synology is selling these parts with ECC UDIMMs, not ECC RDIMMs and in single-channel memory configurations. The Intel Xeon D-1500 series is capable of up to 128GB of memory in dual channel memory configurations with 32GB ECC RDIMMs. 128GB may seem like too much for a storage server, but it seems like a missed opportunity to add higher-quality memory to its enterprise solution.
Networking on the three units is quad 1GbE or quad 1GbE plus dual 10Gbase-T. If one looks at the 2021 generation of servers, 25GbE will be commonplace. Even in 2020 STH was using 25GbE on many servers just because it is a nominal cost upgrade over 10GbE and 25GbE enterprise switch port is growing so quickly. A great example is that in our recent HPE ProLiant DL380T Gen10 Trusted Supply Chain Server Teardown we effectively had the lowest-end configuration that HPE could put together but that even had a dual 10/25GbE NIC.
Between the processors, memory, and networking, the new Synology units seem like what would have been deemed “safe” or “mature” technology in 2017 or 2018. In 2021, this hardware feels almost a generation behind.
For Synology to break into the Enterprise market, we would have liked to see the company do something interesting from a hardware standpoint that was forward-looking into 2021 rather than safe in 2017-2018. Synology has large hardware margins given its pricing which makes sense given its software and warranty. At the same time, there was room to move to a platform that is more modern than state-of-the-art in 2015.
Discussing the Specs: Locking Hard Drives
Perhaps the biggest topic of discussion is locking the three new enterprise units to a compatibility list of hard drives. In the consumer/ SMB space, Synology would be directly going against customer expectations if they were to implement this. In the enterprise space, this makes a lot of sense.
An enterprise storage product is designed to be a package. The software, system, and drives need to be qualified and supported by a vendor. If something goes wrong, enterprises expect to turn to the storage vendor and get an answer. If vendors supported any drive into systems, then it would be nearly impossible to test and it would create a situation where the system vendor could point to the drive vendor and the drive vendor could point to the system vendor.
In order to move into the market, Synology needed to do some type of enforcement. Having a software lock that prevents other types of storage from being used is perhaps extreme, but it is how Synology is dealing with this transition.
Many enterprise vendors effectively do this. There are Seagate and HGST drives out there with special model numbers for large storage vendor arrays. The difference is that Synology is replacing the hard drive vendor label with its own instead of just changing a model number. Synology does support non-Synology drives in low capacities, but not supporting more modern densities is a signal of where this is heading.
Perhaps the biggest challenge at this point is that Synology is using only Toshiba drives, and has very limited capacity points. That is far from a robust supply chain and limits Synology buyers to relatively small capacities. While Seagate/WD are offering 18TB/20TB models, Synology has announced a 16TB model but only has 12TB available now. That means that competitive enterprise platforms can be 50%+ denser than what Synology can ship today in its new Enterprise units.
Hopefully, this is something that expands to multiple vendors and multiple capacity points in the future.
What started out as a “news” post ended up encompassing analysis and even a video from Patrick our Editor-in-Chief. We have received a lot of questions about these Synology announcements recently, so we wanted to get into it in more detail.
We hope going into this level of detail helps our readers understand what Synology is doing, and why. We also hope it clears up some of the questions around the hardware. STH did most of its Xeon D-1500 series coverage before I started writing, but that coverage is one of the reasons I gravitated towards the site.