The Dell PowerSwitch S5248F-ON is a 1RU top-of-rack switch with forty-eight 25GbE SFP28 ports, four 100GbE QSFP28 ports and two 200GbE QSFP-DD ports, a 2.0 Tbps switching fabric (4.0 Tbps full duplex) and 847 nanosecond port-to-port latency - the lowest figure in the S5200-ON family. It ships as an ONIE switch running Dell SmartFabric OS10, and it is the only model in the family that carries the double-density QSFP28-DD cage.
This page is written for the engineer building or expanding a 25GbE leaf layer. Everything below is drawn from Dell's own Installation Guide, the SmartFabric OS10 User Guide, the SmartFabric Services guide, the BMC guide and the transceiver matrix, rather than from the marketing sheet alone - including the port-group rules that decide what you can actually configure, and the licensing mechanics that decide what a second-hand unit is worth.
The port map, in the order the switch numbers them
Dell's I/O-side figure and the OS10 show port-group output agree on the
layout. Reading the front panel left to right: the Stack ID display and status LEDs, then
forty-eight SFP28 cages in two rows of twenty-four, then the two QSFP28-DD cages stacked
vertically, then the four QSFP28 cages in a two-by-two block.
Two things in that numbering matter operationally. First, the QSFP28-DD cages come before the QSFP28 cages in the interface numbering: ports 49 and 50 are the two logical halves of the first DD cage, ports 51 and 52 the second, and only then do the four plain QSFP28 uplinks appear as ports 53 through 56. Scripts written against a conventional 48-port leaf, where port 49 is the first uplink, will land on the wrong cage.
Second, the management ports are not on the port face. On the S5248F-ON the out-of-band Ethernet management port and the RJ45 console port sit on the PSU side of the chassis, together with the USB Type-A port, the reset button and the pull-out luggage tag. Only the S5212F-ON and S5296F-ON put management on the I/O side. If you are planning cable management around a rack where the switch faces the cold aisle, the console and management drops have to reach the opposite face.
Breakout, port groups, and the constraint that catches people
The S5200F-ON series does not use the plain interface breakout command that
most Dell switches use. It uses port groups, and OS10 states the consequence
plainly: when you convert a port to a particular mode, every other port in that group changes
with it. Convert Ethernet 1/1/1 to 10g-4x and interfaces 1/1/2, 1/1/3 and 1/1/4
follow. Speed on the access ports is therefore set in blocks of four, not per port.
That single rule governs most real-world planning on this switch. A rack with a mix of 10GbE and 25GbE servers cannot be wired arbitrarily; the 10GbE hosts have to be grouped into whole four-port blocks, or you waste ports. It is worth drawing your patching plan against the port-group table before the switch is racked.
| Port group | Ports | Supported modes |
|---|---|---|
| 1/1/1 - 1/1/12 | 1-48 (SFP28), four ports per group | 25g-4x, 10g-4x |
| 1/1/13 | 49, 50 (first QSFP28-DD cage) | 100g-2x, 50g-4x, 40g-2x, 25g-8x, 10g-8x |
| 1/1/14 | 51, 52 (second QSFP28-DD cage) | 100g-2x, 50g-4x, 40g-2x, 25g-8x, 10g-8x |
| 1/1/15 - 1/1/18 | 53, 54, 55, 56 (QSFP28), one port per group | 100g-1x, 50g-2x, 40g-1x, 25g-4x, 10g-4x |
Note what the access groups do not offer: there is no 50GbE or 100GbE mode on ports 1 to 48, and no 1GbE breakout mode. The four QSFP28 uplinks, by contrast, are individually addressable - each is its own port group, so you can run one at 100GbE, break the next into four 25GbE and set a third to 40GbE without disturbing the others.
What one QSFP28-DD cage fans out to
The DD cages are the interesting part. Each is a single faceplate opening carrying two
QSFP28 signal groups, and OS10 exposes five modes for it. The default is
100g-2x: two independent 100GbE interfaces from one cage.
Working the arithmetic across the whole chassis from that table gives the real ceilings: eighty ports of 25GbE (48 native SFP28, plus 8 from each DD cage, plus 4 from each QSFP28), the same eighty at 10GbE, eight ports of 100GbE (four native QSFP28 plus two from each DD cage), sixteen at 50GbE and eight at 40GbE. Dell's spec sheet publishes the same 80 / 80 / 8 / 16 figures; its 40GbE row is the one place the two documents disagree, and the OS10 port-group table is the one that reflects what the CLI will accept.
One caution worth stating plainly: a standard QSFP28 module will seat in a DD cage and work, but it only occupies half of it. You get 100GbE out of a 200GbE port. To use the whole cage you need a DD transceiver or a DD breakout assembly, covered below.
What QSFP28-DD actually is, and why the uplink budget is unusual
Dell's transceiver documentation defines the double-density cage precisely: QSFP28-DD modules "transport two 100GbE data streams while consuming the same face plate area as a single 100GbE QSFP28 module." Eight 25G lanes, 200GbE, in one opening.
That is why this switch's uplink budget does not look like a normal 48-port leaf. Four QSFP28 at 100GbE plus two QSFP28-DD at 200GbE is 800 Gbps of uplink against 1.2 Tbps of 25GbE access - a 1.5:1 oversubscription ratio, where a conventional 48-port leaf with four 100GbE uplinks lands at 3:1. Dell's own capacity figure confirms the reading: 48 x 25 + 4 x 100 + 2 x 200 = 2,000 Gbps, exactly the 2.0 Tbps switching capacity quoted. Read the DD ports as 100GbE each and the arithmetic falls short at 1.8 Tbps.
For a leaf carrying vSAN, iSCSI, NFS or RDMA traffic, that extra headroom is the whole argument for this model over a cheaper 48-port alternative. East-west storage replication is exactly the traffic pattern that punishes a 3:1 leaf.
Optics and cables, per port type
Every port type on this switch takes a different family of parts. Cross-referenced against Dell's transceiver and cable matrix:
25GbE access - ports 1 to 48 (SFP28)
Optical modules are SFP28-25G-SR (up to 100 m over parallel multimode), SFP28-25G-ESR (extended short reach, up to 300 m) and SFP28-25G-LR (10 km single-mode). For in-rack copper, DAC-SFP-25G-xM comes in 1, 2, 2.5, 3 and 5 metre lengths and the 1, 2 and 3 metre variants can operate without FEC - a detail worth knowing, because forward error correction is the usual reason a 25GbE link refuses to come up. Active optical AOC-SFP-25G-xM runs 2 to 20 metres and operates with FEC.
OS10 exposes FEC as fec {CL74-FC | CL91-RS | CL108-RS | off}, with the default
determined by the inserted media type: 25G and 50G interfaces use off, CL108-RS or
auto-negotiation, and 100G interfaces use off, CL91-RS or auto-negotiation. When a 25GbE link
comes up on one vendor's NIC and not another's, FEC mismatch is the first thing to check.
10GbE and 1GbE on the same cages
The SFP28 cages are backward compatible, but the path is specific. Set the port group to
10g-4x and the four member interfaces run at 10GbE; from there, the interface
accepts speed 1000 for gigabit. Dell's own worked example runs exactly that
sequence - port-group 1/1/1, mode Eth 10g-4x, then
interface ethernet 1/1/1:1, speed 1000. There is no 1GbE port-group
mode; 1GbE lives underneath 10g-4x. Ten-gigabit copper and optical parts are the standard
DAC-SFP-10G-xM (0.5 to 7 m) and AOC-SFP-10G-xM (2 to 15 m).
100GbE uplinks - ports 53 to 56 (QSFP28)
Point-to-point copper is DAC-QSFP-100G-xM at 0.5 to 5 metres; active optical is
AOC-QSFP-100G-xM at 3 to 30 metres. For breakout into the 25GbE access layer,
DAC-QSFP-4SFP28-25G-xM (2, 3, 5 m) and AOC-QSFP-4SFP28-25G-xM (10 m) split one QSFP28 into
four SFP28 - the physical counterpart of the 25g-4x mode. If you only need a
single low-speed link out of a QSFP28 cage, a QSA28 pluggable adapter lets you seat one SFP28
or SFP+ optic in it, at the cost of the other three lanes.
200GbE double-density - ports 49 to 52 (QSFP28-DD)
The optical module is Q28DD-200G/80G-2SR4 on an MPO-12-DD connector at 850 nm, rated 70 m
over OM3 and 100 m over OM4 at 5.0 W. Dell notes it is compatible with Q28-100G-SR4-G2 or
QSFP+-40G-SR4 for 1x2 breakout, and with SFP28-25G-SR-G2 or SFP+-10G-SR for 1x8 breakout -
mapping directly onto the 100g-2x and 25g-8x/10g-8x
modes. On copper and AOC:
- DD to DD, full 200GbE: DAC-Q28DD-200G-xM (0.5 to 3 m), AOC-Q28DD-200G-xM (5, 10, 20 m)
- DD to 2 x QSFP28 (100g-2x): DAC-Q28DD-2Q28-100G-xM (2, 3 m), AOC-Q28DD-2Q28-100G-xM (5, 15 m)
- DD to 8 x SFP28 (25g-8x): DAC-Q28DD-8S28-25G-xM (2, 3 m), AOC-Q28DD-8S28-25G-xM (10 m)
- DD to 8 x SFP+ (10g-8x): AOC-Q28DD-8SFP-10G-xM (10 m)
The DD breakout parts are the least common items in this list on the secondary market, and they are the ones that unlock the switch's advantage. Budget for them deliberately rather than assuming a generic QSFP28 cable will do the job.
SmartFabric OS10 licensing on the secondary market
This is the single most important thing to understand before buying any ONIE switch second-hand, and it is more favourable than most buyers expect - with one specific trap.
Dell's position is documented: a factory-loaded OS10 image includes a perpetual
license, pre-installed on the switch as a licence file bound to the service tag and
stored at /mnt/license/<service-tag>.lic. If you pre-ordered a switch with
OS10, you never install a licence at all. You can verify the state of any unit in one command,
show license status, which reports the software edition, version, License Type,
License Duration and License Status. A healthy factory unit reads PERPETUAL, Unlimited,
Active.
The trap is what happens when someone wipes the switch. Dell states it directly: if a factory-installed OS10 image is deleted after booting into ONIE Install OS mode, you must re-install the licence. Recovering it means signing in to the Dell Digital Locker under the purchaser's account, locating the entitlement ID and order number, and confirming the device service tag appears in the Assigned To field. A second owner generally does not have access to that account. In other words: an ONIE "uninstall OS" performed as a well-meaning data wipe can permanently separate a switch from its perpetual licence.
An OS10 image downloaded rather than factory-loaded carries a 120-day trial
licence. When it lapses the switch keeps forwarding, but logs
%LADF_LIC_EVAL_PRD_NOTICE to the console once an hour until a perpetual licence is
installed.
What that means in practice, stated honestly: we sell the hardware. We do not promise a
transferred Dell entitlement, a software subscription or a support contract with a refurbished
switch, and you should treat any reseller who does with suspicion. What we will do is tell you
the state of a specific unit before it ships - whether OS10 is present, what
show license status reports, and what version it is on. If you would rather own
the software stack yourself, the chassis is indifferent: ONIE will install Enterprise SONiC
Distribution by Dell Technologies, community SONiC, or another ONIE-compatible network
operating system. Plan the NOS as its own line item either way.
Recovering a switch that arrives at the ONIE prompt
If a unit boots to bare ONIE, the mechanics are straightforward. The Grub menu offers ONIE
Install OS, Rescue, Uninstall OS, Update ONIE, Embed ONIE and EDA-DIAG. ONIE service discovery
tries locally attached storage first, then TFTP via DHCP, then IPv6 link-local neighbours over
HTTP, and you can stop it with onie-discovery-stop and install manually with
onie-nos-install <URL> over HTTP, FTP or TFTP. The install environment takes
a DHCP address on the management interface; if DHCP fails it falls back to
192.168.3.10/255.255.255.0, which is the address to point a laptop at when
nothing else responds. Remove any recovery USB stick before running
onie-nos-install. USB storage is FAT and does not auto-mount - you mount it by
hand from ONIE rescue mode.
SmartFabric mode versus Full Switch mode
OS10 runs the switch in one of two personalities, and Dell is explicit that PowerSwitches boot in Full Switch mode by default. Full Switch is the conventional CLI: you configure interfaces, VLANs, BGP and VLT yourself.
SmartFabric mode hands fabric construction to SmartFabric Services. You enable it with
smartfabric l3fabric enable role LEAF vlti ethernet 1/1/4-1/1/5, confirm the
reboot, and the switch comes back with the CLI deliberately restricted to global switch
management and monitoring - the fabric itself is then driven from the SFS GUI or REST API.
show switch-operating-mode confirms which personality is active, and
no smartfabric l3fabric deletes the fabric configuration and reboots back into
Full Switch mode. Enabling SFS on a leaf pair automatically configures the two switches as a
VLT pair.
The S5248F-ON is a Dell-validated leaf role in SFS, offering 10GbE or 25GbE node connectivity; the S5232F-ON is the matching spine. An SFS fabric scales to twenty switches across eight racks, leaves must be deployed in pairs, and both members of a pair must be the same model because of the VLT requirement. Two further constraints are worth knowing before you cable: breakout is not supported on leaf switches for the VLTi or leaf-to-spine links outside PowerScale back-end deployments, so use native-speed cables there, and autobreakout - enabled by default on spine switches - only works on DAC.
VLT as deployed on this platform
Virtual Link Trunking is how you get an active-active pair with no spanning-tree blocked link. OS10 supports VLT port-channel termination across two switches, so a pair of S5248F-ON units presents ninety-six active 25GbE server ports in two rack units. The details that decide whether a build works:
- VLAN 4094 is reserved as the internal control VLAN for the VLT domain, and uses IPv6 addressing for peer-to-peer communication. Do not allocate it.
- Both peers must run the same OS10 version, with a temporary exception during upgrades. On the used market that means buying a matched pair, or budgeting to bring both units to a common release.
- The primary node is the one with the lowest MAC address by default; you
override that with
primary-priority. If the primary fails and later returns it comes back as secondary - there is no preemption. - ARP, IPv6 neighbour and MAC tables synchronise between peers, and VRRP Active-Active is enabled by default on all VLAN interfaces so each peer forwards Layer 3 locally instead of hairpinning across the VLTi.
- A domain ID mismatch between peers stops the VLTi coming up altogether.
- 802.1x and DHCP snooping are not supported on VLT ports. RSTP, RPVST+, MSTP, IGMP snooping and MLD snooping are.
One genuinely useful nuance for a phased upgrade: OS10 allows switches of the same group and the same port media type to share a VLT domain, and names S5224F-ON plus S5248F-ON as a valid combination. So a 24-port leaf can be paired with a 48-port leaf under VLT even though SmartFabric Services would require identical models. Mixed media types - the example Dell gives is S4148F-ON with S4148T-ON - cannot share a domain.
Airflow, power supplies and fans
Airflow direction is set at the factory by SKU, not chosen in the rack. Dell builds the series in four combinations: AC or DC PSU with fan airflow from the I/O side to the PSU side (normal), and AC or DC PSU with airflow from the PSU side to the I/O side (reverse). Crucially, Dell gives you a way to identify them by eye: the red indicator is normal airflow and the blue indicator is reverse airflow. On a used unit with an ambiguous listing, that colour is the fastest check there is.
Which direction is correct is decided by which way the switch faces. In a hot-aisle/ cold-aisle row, servers draw cool air from the cold aisle and exhaust into the hot aisle. A switch racked with its port face toward the cold aisle needs normal airflow. A switch racked with its port face toward the hot aisle - a very common choice, because it puts the switch ports on the same side as the server NICs and halves the cable runs - needs reverse airflow, or it will inhale its own row's exhaust. Dell is unambiguous that you must not mix the two inside one chassis: all fans must use the same airflow direction, and a mixed installation has to be corrected to avoid damage.
Two maintenance facts change how you should think about spares. First, each power supply has an integrated fan that cannot be replaced on its own - if the PSU fan fails, the whole PSU is replaced. The four fan trays are separately field-replaceable. Second, a fan module swap has to be completed within one minute or the switch temperature can rise past safe thresholds and shut it down. Dell also recommends checking the fans at six-month intervals and monitoring fan speeds to set replacement intervals.
The chassis takes two hot-swappable power supplies in a 1+1 arrangement, and will run on one: install the single PSU in the PSU1 slot and fit a blank plate in PSU2. Status is straightforward at the LED level - solid green for input OK, flashing yellow for a PSU fault, flashing green at 1 Hz for standby, off for no power; fans show solid green for normal and flashing yellow for a fault. A DC variant exists, wired through a terminal block where blue is -48V, black is positive return and green/yellow is ground.
Because power supplies and fan modules are direction-specific parts, the direction is fixed when the chassis is built rather than changed in the rack. Tell us which direction your row needs and we will confirm the build before anything ships.
Racking, depth and the console details nobody documents
The chassis is 17.1 inches wide, 18.1 inches deep and 1.72 inches high (434 x 460 x 43.6 mm), weighing 21.4 lbs with power supplies and fans fitted. Dell specifies five inches of clearance front and back, and a minimum of five inches between intake and exhaust vents and the cabinet wall, so plan on roughly 28 inches of usable rack depth rather than 18. If you install two of these near each other, Dell asks for five inches between them for proper airflow.
Mounting uses the ReadyRails system, which the S5248F-ON shares with the S5224F-ON and S5232F-ON but not with the S5296F-ON or S5212F-ON. ReadyRails supports four installations: 1U tool-less mounting in four-post square or round-hole racks, two-post flush-mount, two-post centre-mount, and four-post threaded. Only the tool-less method needs no screws; the other three are tooled, and Dell does not include rack screws in the box. The rails engage a locking feature roughly three inches before the switch is fully inserted. A ground lug kit is included, and the chassis takes a two-hole ground lug.
Console access has two genuinely obscure gotchas. The RJ45 console runs at 115200 baud, no parity, eight data bits, one stop bit, no flow control. But Dell warns that while a console cable connected through to a patch panel or terminal server can run the full 100 m, a cable plugged into the console port and left disconnected at the far end must be under 6 metres - if it is longer, the switch may not boot. A dangling console tail is a genuine cause of a switch that will not start. Separately, the MicroUSB-B console uses a Silicon Labs CP2102 bridge, so you need the SiLabs CP210x USB-to-UART driver on the laptop, not a Dell driver. When the MicroUSB-B port is connected it becomes the primary console and all messages go there.
Identification matters on the used market. The S5248F-ON is regulatory model
E21W, regulatory type E21W002 - the S5232F-ON is E21W005 and
the S5224F-ON is E21W003 on the same regulatory model, so the type suffix is what separates
three physically similar 1RU switches. The pull-out luggage tag on the PSU side carries the
service tag, MAC address, PPID and express service code; the service tag is also what
show chassis reports and what any licence file is named after.
The BMC, and reading FRU data without a network OS
The S5200-ON series carries a real IPMI baseboard management controller, documented in its
own guide and largely unmentioned in the marketing material. It is reachable with
ipmitool over the network using the lanplus interface, and it exposes
system health sensors, host power control, firmware update, the system event log and - most
usefully for an inbound refurbished unit - FRU inventory. ipmitool fru print
enumerates the builtin FRU device plus FRU_PSU1, FRU_PSU2 and the fan FRUs, each reporting
board manufacturer, manufacture date, board serial and board part number.
That is the reliable way to establish exactly which power supplies and fan modules are
physically installed in a given chassis, and it works without a licensed network operating
system. Two security notes follow from it: Dell's examples use admin/admin
credentials over LAN, and accessing the BMC from the host requires no username or password at
all, so changing the BMC credentials and confirming the preprogrammed static BMC address should
be on the commissioning checklist for any second-hand unit.
Environmental limits worth checking against your site
Operating range is 0 to 45 C (32 to 113 F) continuous, and the switch is Dell Fresh Air 2.0 compliant to 45 C. Two constraints get overlooked: the maximum temperature must be reduced by 1 C for every 125 metres above 950 metres of altitude, and maximum operational altitude is 10,000 feet. A site at 6,000 feet is giving up roughly 12 C of thermal headroom before anything else happens. Maximum thermal output is 647W, or 2,208 BTU/h, against 310W typical draw - so a pair in a VLT domain is a little over 4,400 BTU/h at worst case for rack cooling budgets.
Choosing within the S5200-ON family
All five models share the same operating system, the same 32 MB packet buffer and the same Layer 2 and Layer 3 feature set. You are choosing density and uplink ratio, not capability.
| Model | Ports | Where it fits |
|---|---|---|
| S5212F-ON | 12x 25GbE SFP28 + 3x 100GbE QSFP28 | Half-width 1RU; hyper-converged and small edge racks; fixed PSUs and fans |
| S5224F-ON | 24x 25GbE SFP28 + 4x 100GbE QSFP28 | 1RU; low-density server and storage aggregation. Can share a VLT domain with this switch |
| S5248F-ON (this switch) | 48x 25GbE SFP28 + 4x 100GbE QSFP28 + 2x 200GbE QSFP28-DD | 1RU; the mainstream 25GbE leaf, the only model with QSFP28-DD, and the lowest latency in the family at 847 ns |
| S5296F-ON | 96x 25GbE SFP28 + 8x 100GbE QSFP28 | 2RU; middle-of-row and end-of-row aggregation, highest 25GbE density. Does not use ReadyRails |
| S5232F-ON | 32x 100GbE QSFP28 + 2x 10GbE SFP+ | 1RU; the 100GbE spine for this family, and the SFS spine role. Uses interface breakout rather than port groups |
Choose the S5224F-ON if a rack will never exceed twenty-four hosts, the S5296F-ON if you are consolidating two racks into one end-of-row switch and can spend 2RU, and the S5232F-ON if the job is spine rather than leaf. Choose this model when you need forty-eight 25GbE access ports with uplink headroom that a four-port-100GbE leaf cannot give you. Stepping up a generation, the Dell PowerSwitch S5448F-ON moves the same design language to 100GbE access with 400GbE uplinks.
Common questions
- Are the QSFP28-DD ports 200GbE or 100GbE?
- 200GbE each. Dell's Installation Guide lists the S5248F-ON as "forty-eight 25GbE SFP28 ports, four 100GbE QSFP28 ports, and two 200GbE QSFP-DD ports", the ordering description reads "2x 200GbE QSFP28-DD", and OS10 shows both DD port groups defaulting to
100g-2x- two 100GbE interfaces per cage. One page of the spec sheet's technical specifications prints "2x100GbE QSFP28-DD"; that reading contradicts Dell's own 2.0 Tbps capacity figure, which only balances at 200GbE per cage. - How many 100GbE ports does it really have?
- Eight. Four native QSFP28 on ports 53 to 56, plus two from each QSFP28-DD cage in
100g-2xmode. That is twice the 100GbE count of a conventional 48-port 25GbE leaf. - Can I run some access ports at 10GbE and others at 25GbE?
- Yes, but only in blocks of four. Ports 1 to 48 are organised into twelve port groups of four, and changing the mode on one member changes all four. Plan the patching so 10GbE hosts fill whole groups.
- Will it do 1GbE?
- Yes, underneath 10GbE. Put the port group into
10g-4xmode, then setspeed 1000on the individual interface. There is no dedicated 1GbE port-group mode. - Does a refurbished unit come with an OS10 licence?
- A factory-loaded OS10 image includes a perpetual licence tied to the service tag, and many used units still carry it -
show license statuswill say PERPETUAL / Active. But if a previous owner ran ONIE Uninstall OS, the licence file goes with it and recovery requires the original purchaser's Dell Digital Locker account. We will tell you what state a specific unit is in, and we do not promise a transferred entitlement or a support contract. - How do I tell normal airflow from reverse on a used switch?
- By the indicator colour on the power supply and fan modules: red is normal (I/O panel to PSU), blue is reverse (PSU to I/O panel). Airflow is set at the factory by SKU and the parts are direction-specific, so all four fans and both PSUs must match.
- What is the oversubscription ratio?
- 1.5:1 - 1.2 Tbps of 25GbE access against 800 Gbps of uplink. The headroom comes from the QSFP28-DD cages and is the main reason to choose this model for storage or RDMA traffic.
- Where are the management ports?
- On the PSU side, not the port face. The out-of-band Ethernet management port sits above the RJ45 console port, alongside the USB Type-A port, reset button and luggage tag. Console is 115200 8N1, no flow control.
Condition, testing and what ships in the box
This is a refurbished, professionally reconditioned unit. Every switch is powered up, inspected and functionally tested before it leaves us, and it ships with warranty coverage. Power supplies, fan modules, rail kits, console cables and optics are configurable above - choose the airflow direction that matches your row, because power supplies and fans are direction-specific parts and cannot be mixed within one chassis.
If you need a specific firmware or NOS state, a particular optics loadout, DD breakout assemblies, or matched pairs at a common OS10 release for a VLT domain, tell us before you order and we will build to it.
Specifications
| Model | S5248F-ON |
|---|---|
| Regulatory Model / Type | E21W / E21W002 |
| Ports | 48x 25GbE SFP28, 4x 100GbE QSFP28, 2x 200GbE QSFP28-DD |
| Interface Numbering | 1-48 SFP28; 49-52 QSFP28-DD (two logical ports per cage); 53-56 QSFP28 |
| Uplinks | 4x 100GbE QSFP28, 2x 200GbE QSFP28-DD |
| Interface / Media | SFP28 (1/10/25GbE) / QSFP28 (10/25/40/50/100GbE) / QSFP28-DD (200GbE double-density) |
| Connector Types | SFP28, QSFP28, QSFP28-DD |
| Data Rates | 25GbE access (SFP28); 100GbE (QSFP28); 200GbE (QSFP28-DD) |
| Port Groups | 1/1/1-1/1/12 = ports 1-48 in blocks of four (25g-4x, 10g-4x); 1/1/13-1/1/14 = QSFP28-DD cages (100g-2x, 50g-4x, 40g-2x, 25g-8x, 10g-8x); 1/1/15-1/1/18 = QSFP28 (100g-1x, 50g-2x, 40g-1x, 25g-4x, 10g-4x). Mode change applies to every port in the group |
| Switching Capacity | 2.0 Tbps |
| Switch Fabric (Full Duplex) | 4.0 Tbps (full duplex) |
| Forwarding Performance | 1.5 Bpps (3.0 Bpps full duplex) |
| Port-to-Port Latency | 847 ns - lowest in the S5200-ON family |
| Packet Buffer Memory | 32MB |
| CPU | Four-core Intel Denverton |
| CPU Memory | 16GB |
| Onboard Storage | 64GB SSD |
| Rack Units | 1U |
| Dimensions | 1.7 x 17.1 x 18.1 in (H x W x D); 43.6 x 434 x 460 mm |
| Rack Clearance Required | 5 in (12.7 cm) front and back; 5 in between adjacent units |
| Rail System | ReadyRails 1U - tool-less four-post, two-post flush-mount, two-post centre-mount, four-post threaded. Rack screws not included. Ground lug kit included |
| Weight | 21.4 lbs (9.7 kg) with PSUs and fans |
| Power Supplies | 2x hot-swappable 750W 80 PLUS Platinum 100-240 VAC AC power supplies (DC available as a custom kit). Each PSU has an integrated fan that is not separately replaceable |
| Power Redundancy | Yes - 2x hot-swappable AC power supplies (1+1). Single-PSU operation supported in the PSU1 slot with a blank plate in PSU2 |
| Maximum Power Consumption | 647W maximum |
| Typical Power Consumption | 310W typical |
| Maximum Current Draw | 5.8A @ 110VAC / 2.9A @ 220VAC per the S5200 spec sheet; the Setup Guide lists 2.4A @ 220VAC |
| DC Power (optional build) | -40VDC minimum input; 16.2A at 40VDC maximum at full load with fan |
| Maximum Thermal Output | 2208 BTU/h |
| Fan Modules | 4x hot-swappable fan modules; slots 1-2 left, 3-4 right. Replacement must be completed within one minute |
| Airflow Direction | Factory-set by SKU. I/O Panel to PSU (normal, red indicator) or PSU to I/O Panel (reverse, blue indicator). Directions must not be mixed within a chassis |
| Operating System | Dell SmartFabric OS10; also supports Enterprise SONiC Distribution by Dell Technologies and alternative network operating systems via ONIE |
| OS10 Licensing | Factory-loaded OS10 includes a perpetual licence bound to the service tag at /mnt/license/<service-tag>.lic; verify with show license status. Downloaded images carry a 120-day trial licence |
| Operating Modes | Full Switch mode by default; SmartFabric mode via smartfabric l3fabric enable. Validated SFS leaf role (S5232F-ON is the spine); SFS scales to 20 switches across 8 racks |
| Switching / Routing | Layer 2 and Layer 3 (OSPF, BGP, PBR, IPv4 and IPv6) |
| Multi-Chassis | Virtual Link Trunking (VLT) and Routed VLT. Two-node domains; VLAN 4094 reserved for VLT control; peers must run the same OS10 version; no primary preemption; may pair with S5224F-ON |
| Forward Error Correction | CL74-FC and CL108-RS (25G/50G), CL91-RS (100G), or off. Default set by inserted media type |
| QoS | QoS with Data Center Bridging: priority flow control (802.1Qbb), enhanced transmission selection (802.1Qaz), DCBx, iSCSI TLV |
| Overlay | VXLAN gateway bridging and routing at line rate; BGP EVPN with Integrated Routing and Bridging (asymmetric and symmetric) under Dell SmartFabric OS10 |
| RDMA | Routable RoCE |
| Precision Time Protocol | IEEE 1588v2 PTP in hardware under Dell SmartFabric OS10 |
| MAC Addresses | 32K min, 288K max (ALPM dependent) |
| IPv4 Scale | 16K min / 168K max hosts; 128K routes |
| IPv6 Scale | 8K min / 100K max hosts; 64K routes |
| Multicast Routes | 16K |
| VLANs | 4K (VLAN 4094 reserved when VLT is configured) |
| MSTP Instances | 63 |
| PVST Instances | 150 |
| Total LAG | 128 (16 members per LAG) |
| ACL Scale | L2 ingress 2K / L2 egress 256 / IPv4 ingress 2K / IPv4 egress 2K / IPv6 ingress 1K / IPv6 egress 1K |
| Maximum 10GbE Density | 80 - 48 (SFP28) + 16 (QSFP28-DD breakout) + 16 (QSFP28 breakout) |
| Maximum 25GbE Density | 80 - 48 (SFP28) + 16 (QSFP28-DD breakout) + 16 (QSFP28 breakout) |
| Maximum 40GbE Density | 8 - 4 (QSFP28 at 40g-1x) + 4 (QSFP28-DD at 40g-2x) per the OS10 port-group table; the S5200 spec sheet lists 6 + 4 |
| Maximum 50GbE Density | 16 - 8 (QSFP28-DD at 50g-4x) + 8 (QSFP28 at 50g-2x) |
| Maximum 100GbE Density | 8 - 4 (QSFP28) + 4 (QSFP28-DD breakout) |
| Management | 1x RJ45 console port with RS232 signaling (115200 8N1, no flow control), 1x MicroUSB-B console (Silicon Labs CP2102 bridge), 1x 10/100/1000BaseT out-of-band Ethernet management port, 1x USB Type-A. All located on the PSU side |
| Console Cable Limit | 100 m when terminated at a patch panel or terminal server; under 6 m if left disconnected at the far end, or the switch may not boot |
| Out-of-Band Hardware Management | IPMI baseboard management controller - sensors, FRU inventory (ipmitool fru print), host power control, SEL and firmware update over lanplus |
| Zero-Touch Provisioning | Open Network Install Environment (ONIE); ZTD enabled by default on factory-installed OS10. ONIE fallback management address 192.168.3.10/255.255.255.0 |
| Identification | Pull-out luggage tag on the PSU side carrying service tag, MAC address, PPID and express service code |
| PoE | No |
| Form-Factor | 1RU |
| Input Power | 100-240 VAC, 50/60 Hz |
| Operating Temperature | 32 to 113 F (0 to 45 C) continuous; reduce maximum by 1 C per 125 m above 950 m |
| Operating Humidity | 5 to 90% RH, non-condensing (Installation Guide states 5 to 85% RH continuous) |
| Storage Temperature | -40 to 158 F (-40 to 70 C) |
| Maximum Altitude | 10,000 ft (3,048 m) operational; 39,370 ft (12,000 m) non-operational |
| Fresh Air | Dell Fresh Air 2.0 compliant to 45 C |
| Regulatory | UL/CSA 60950-1 2nd Ed., EN 60950-1, IEC 60950-1, EN 60825-1/2, FDA 21 CFR 1040.10 and 1040.11; FCC CFR 47 Part 15 Subpart B Class A; EU RoHS compliant |
| Certifications | Available with US Trade Agreements Act (TAA) compliance; USGv6 Host and Router certified; IPv6 Ready (Host and Router); UCR DoD APL |
