Solana ShredStreamneeds a server to land on
A shred feed is only as fast as the machine receiving it. We are the server side of that setup: single-tenant bare metal at Jito-connected edges, from $299.99/mo, built to absorb a UDP firehose without dropping packets.
The feed itself comes from OrbitFlare ShredStream, which delivers raw shreds over UDP with no proxy and no keypair. Take both in the same city and the head start survives the hop.
Unlimited-bandwidth 10Gbps ports are verified for Frankfurt and Amsterdam. US locations are metered with a generous transfer allowance.
How the pieces fit
A feed, a machine, and the distance between them
Consuming shreds takes two purchases and one decision. Buy the feed from the people who operate it, buy the machine from the people who operate the racks, and put both in the city where your strategy already runs.
The feed
OrbitFlare ShredStream
Raw shreds over UDP from top-of-Turbine validators, no proxy and no keypair. 9 regions, slots from $500/mo, free 60-minute trial. OrbitFlare builds and operates it.
See ShredStream slotsThe machine
OrbitServers bare metal
Single-tenant hardware at the same Jito-connected edges: no hypervisor between the NIC and your parser, high-clock AMD silicon, NVMe, dedicated IPv4, full root and IPMI/KVM.
See bare metal configurationsThe path
One city, one network
Take the slot and the server in the same city and the head start survives. Feed, strategy, and submission path stay on one network instead of crossing the public internet twice.
See measured Jito latencyWhy now
The free Jito feed is going away
In July 2026, Jito announced it is deprecating ShredStream: in its own words, “the service will be completely shut down” on September 5, 2026. After that date shredstream-proxy stops authenticating and the shreds stop arriving. Everything downstream of the feed, your parsers, filters, and strategies, keeps working if you swap the source.
Jito’s docs point migrants to DoubleZero Edge. OrbitFlare ShredStream is the other option, and it is the one that runs where our servers do. Whichever feed you land on, the server question is the same: it has to sit in the same city, and it has to keep up with the packets.
The server side
What a shred consumer needs from its machine
Most of the ShredStream conversation is about feeds. This part is ours: the box that receives the packets decides how much of the head start you actually keep. A ShredStream server has one job, which is to take every packet the moment it arrives and hand it to your code without a queue in between.
A port that will not throttle
Shred delivery is a sustained UDP firehose, not a request/response workload. Frankfurt and Amsterdam ship a 10Gbps port with unlimited bandwidth, so a busy feed plus an RPC node beside it never runs into a transfer cap mid-session. US sites are metered with a generous allowance, which is worth modelling before you point a firehose at one.
No hypervisor in the packet path
On a VM, the hypervisor decides when your process gets the CPU. That scheduling jitter lands directly on tail latency, which is the number that decides races. Bare metal removes the layer: your parser is the only thing competing for the core it is pinned to.
Clocks over core count
Deshredding, deduplicating, and filtering is a serial hot path. A high-clock Ryzen 9 or EPYC part does more for first-seen timestamps than a bigger core count, which is why our Frankfurt fleet is built around high-frequency silicon rather than dense many-core chips.
Kernel tuning you actually control
UDP receive buffers, busy polling, IRQ affinity, CPU pinning, NUMA placement, power governors. Full root access means you tune the box for packet capture instead of filing a ticket, and IPMI/KVM means a bad NIC setting is a console away from being fixed rather than a support queue.
A stable IP to deliver to
Feeds are delivered to an address you nominate, so the destination has to be yours and it has to stay put. Every server includes a dedicated IPv4 with DDoS protection on the network side, and additional addresses are available if you split capture from execution.
Room for the rest of the stack
The shred consumer is rarely alone on the box. Leave headroom for the RPC or Geyser node you reconcile against, the database that holds state, and the strategy processes themselves. Up to 1TB of ECC memory and multiple NVMe drives per chassis means one machine can hold the whole loop.
Unlimited-bandwidth details are on the unmetered 10Gbps page, and the hardware itself is listed on bare metal servers. If you would rather describe the workload and let us size it, talk to sales.
Pick the city first
Servers at every Jito-connected edge
Match the slot region to the server city and the feed never leaves the metro. Latency to Jito is published per location as a measured minimum, average, and maximum, so you can choose on numbers rather than on claims.
Frankfurt bare metal server
London bare metal server
Amsterdam bare metal server
New York bare metal serverSalt Lake City bare metal serverReconciling shreds against confirmed state? See Solana RPC hosting, the per-location Jito latency benchmarks, and the Solana trading stack these servers usually sit inside.
Migrating off Jito
Swapping the source without touching the strategy
The change is source-only, which is why it fits in an afternoon rather than a sprint. Do it while the old feed is still live and the rollback stays free.
- 1
Inventory the consumers
List every process that receives shreds from your shredstream-proxy destinations. Those endpoints are the only thing that changes.
- 2
Run the new feed in parallel
Point the replacement at a second UDP port on the same host. Your existing feed keeps running untouched while you compare.
- 3
Compare p99, not averages
Log first-seen timestamps per signature from both feeds. Tail latency decides races; a better median with a worse p99 is a downgrade.
- 4
Cut over while the old feed is still up
Switch consumers to the new port with the old feed as the live fallback, so a regression is a one-line revert rather than an outage.
- 5
Retire the proxy
Once the new source is carrying production, remove shredstream-proxy and its firewall rules. One fewer process to supervise.
Hold an OrbitFlare NFT Pass? Each active Pass includes one free ShredStream slot (limited time), and delivery is restricted to OrbitServers IPv4 prefixes. The free feed only lands on a server hosted here.
We Accept
All major payment methods
Common Questions
Solana ShredStream FAQ
ShredStream forwards raw Solana shreds, the packets validators exchange while a block is still being assembled, straight to a server over UDP. You see transactions before they reach any RPC or gRPC endpoint, which is the edge trading bots, MEV searchers, snipers, and liquidation engines are built around. Two things make it work: a feed to subscribe to, and a machine close enough to the feed that the head start survives the network hop.
Yes. Jito announced in July 2026 that it is deprecating ShredStream, with the service completely shut down on September 5, 2026. Jito's own docs point migrants to DoubleZero Edge. OrbitFlare ShredStream is the other drop-in option, and it runs at the same Jito-connected edges as our bare metal fleet, so the feed and your strategy can sit in the same city.
From OrbitFlare, which builds and operates the feed. Slots start at $500/mo in standard regions and $1,000/mo in premium regions across nine regions, with a free 60-minute trial you can run beside your current feed. OrbitServers is the server side of that setup: we host the box the feed is delivered to. If you want both arranged together, our sales team will coordinate it.
A machine that can absorb a firehose of UDP without dropping packets. In practice that means bare metal rather than a VM (no hypervisor scheduling jitter between the NIC and your process), high single-thread clocks for the parse path, generous kernel socket buffers, and a port that will not throttle. Our EU sites ship a 10Gbps unmetered port, every server includes a dedicated IPv4 the feed can be pointed at, and you get full root plus IPMI/KVM to tune the stack yourself.
Not with OrbitFlare ShredStream. It delivers standard Solana shreds by UDP straight to the IP and port you nominate, with no proxy process and no keypair authentication. Everything downstream of the feed, your parsers, filters, and strategy code, keeps working unchanged. One fewer daemon to supervise on the box.
The one where your execution already runs. A feed in Frankfurt delivering to a bot in another city reintroduces exactly the latency the feed was bought to remove. Pick the city first, then take the slot and the server there. Every OrbitServers bare metal location is a Jito-connected edge and our per-location Jito latency is published with the measured minimum, average, and maximum rather than a marketing number.
Each active OrbitFlare NFT Pass includes one free ShredStream slot as a limited-time offer, and delivery is restricted to OrbitServers IPv4 prefixes. In other words the free feed only lands on a server hosted with us, which makes a bare metal box at a Jito edge the natural home for it.
Yes, and many teams do. A shred feed tells you what is coming; an RPC or Geyser node tells you what landed. Running both on the same box, or on two boxes in the same rack, keeps that reconciliation loop local. Our Solana RPC hosting pages cover the sizing for the node side.